Title

Even in older stages of the language, numbers are either of Sino-Japanese or native origin. In these older stages, however, two differences are apparent. 

1. The mixing of the two number systems was not as convoluted as is the case today. For instance, the Sino-Japanese numeral for four was understood to be シ, and it is only in recent history that it has been replaced by よん. 
2. The native number system was more productive than it is today and could theoretically create numbers up to 99,999, although it has always been rather impractical past 100. 

※It is thought that when the 呉音 reading サン for 3 first entered the language that it was pronounced as /samu/. 

※The 呉音 reading ク for 9 was standard up until the 20th century when the 漢音 reading キウ supplanted it. 

※As for 10000, the modern 音読み of マン is actually a 慣用音 that appeared much later in history. 
※Sino-Japanese vocabulary contains special Kanji for the numbers 20 (廿・ 廾), 30 (丗・卅), 40 (卌), and 200 (皕). The 呉音・漢音 listed for these characters derive from contractions of the typical two-morpheme construction of these numbers. Meaning, ニジフ, サンジフ, シジフ, and ニヒヤク respectively were always the typical pronunciations for these numbers. 

As hinted at by the previous chart, a few sound changes and irregularities affect native numbers greater than 10. 

• とを undergoes a sound change and manifests as そ - /t/ → /s/ in 30-90.  
• はた meaning 20 is particularly irregular and may be an ablaut form of 2 (ふた). 
• Oddly enough, 50 is simply い in even more ancient language. Proof of this can be found in the surname 五十鈴（いすゞ）meaning "fifty bells." 
• 100 in isolation is もも. As for multiples of 100, it often manifests as ほ, itself being derived from もも. 
• Numbers 100 and greater only gained numerical values centuries after the words themselves appeared in the language, but once they were given their respective values upon influence from the Sino-Japanese number system, they behaved accordingly. 

The generic counter counts things in general and may also indicate age. It manifests in two iterations: ～つ (1-9) and ～ち・ぢ (20+). It is unclear when most if not all instances of ～ち became pronounced as ～ぢ, but with exception to はたち, it is acceptable to read all instances as ～ぢ.

※Notice how 50 is いそ rather than いつそ. This would imply how 5 was originally monosyllabic as い, but this also conflicts with how 50 oddly was also い as mentioned earlier with the surname 五十鈴. 

※The pronunciations もも and ほ for 100, as mentioned above, share the same etymology. For 200-400, it appears that they were interchangeable, although there is actually no recorded instances of some multiples of 100 such as 300. However, given how native number use gets cut off today at 4 (treating もも and ほ as counters), this may have prevented the use of もも for 400 and higher.

※The use of ～ち・ぢ beyond 100 is contested due to sheer lack of instances of numbers that high being used in the native number system. Proof to the unlikelihood of this can be found in the etymology of よろづ, which is thought to derive from or at least be cognate with the Korean word "yeoreo" meaning "various." The ～づ would then be the counter ～つ and indicate that at one point, ～ち・ぢ might not have existed at all. Evidence of this can be found in ～つ attaching to large numbers.

Ex. 五百津真賢木 （いほつまさかき）
Lit. 500 (branched) true-sakaki (sacred trees)
Meaning: Native word for evergreen tree, holding religious significance in Shintoism.

Orthography Note: ～ち・ぢ was given the Ateji spelling of 路. In Modern Japanese, this has caused the multiples of 10 to be misunderstood in the context of age as referring to the entire decade. Thus, 三十路 is often used to mean "one's thirties," but in Classical Japanese, it was understood to mean "age 30."

※Of course, Sino-Japanese numbers have always been paired with Sino-Japanese counters, and the array of such counters that exist today appear in Classical Japanese as well. What would differ would be their pronunciations due to systematic sound changes that have taken place over the centuries.   

Creating numbers in between powers of 10 beyond 10^1 is a murky situation in the native number system. Such means to create any number such as 11, 101, 393, etc. only came into being once speakers felt the system ought to just as productive as the Sino-Japanese system. 

Once native numbers were being used more frequently at least in language devoid of Sino-Japanese vocabulary, speakers would use the native numbers for 1-10, 20, 30, 40, 50, 60, 70, 80, 90, and 100 quite regularly when counting, as is indicated by a passage in the Tale of Genji:

お指（よび）をかがめて十（とを）、はた、みそ、よそ、など数ふるさま
(Her) counting with her fingers 10, 20, 30, 40, etc.

Texts from these initial centuries after the influx of Sino-Japanese influence indicate just how speakers overcame the lack of native numbers in between these base numbers. The solution was the use of the word あまり meaning "remainder."

Take the base unit with its counter, insert あまり, then follow with the remaining part of the number along with the counter once more. Thus, you get numbers like:

In larger numbers, though, a power may or may not have a value. In other words, the difference between 902 and 922 is that the first number has nothing in the tens place whereas the second number does. In this situation, あまり appears between each power with its counter. 

The redundant nature of the native number system is undoubtedly a major factor into its ultimate demise. Before it would fade into history, however, attempts were made to simplify it. 

By the Kamakura Period (12th century), word-medial counters had become simplified to ～ち・ぢ. Thus, みそとせ あまり いつとせ (35 years) would become みそぢ あまり いつとせ.

Then, by the Edo Period, あまり became dropped entirely, but before it took its last breath, it had become common-place to drop all but the last instance. This would render numbers such as 99,999 as:

ここのよろず ここのち ここのほ ここのそぢ あまり ここのつ.

In this example, instances of ～ち beyond 100 are presumed to have not occurred.

Into Early Modern Japanese, any semblance of the native system applied to numbers greater than 10 fundamentally lack あまり. The greatest example of this is 三十一文字, which reads as みそひともじ, referring to the number of letters in a tanka poem. Indeed, in poetry mora count is especially important and would have been a driving factor for this simplification. 

Perhaps the greatest wealth of examples of native numbers can be found in compound expressions. Putting aside whether the number is being used for its numeric value or not, these words have managed to survive into the present as set phrases. 

Common counters other than the generic counter include those for time such as ～年（とせ）for "years," ～月,（つき）for "months," and ～日（か）for "days." 

In the chart below, readings are given for number-counter combinations that would be practical for said counters. Some irregularities do exist, which will be looked at individually afterward. 

※When referring to the first of the month, the traditional reading of 一日 is つきたち, which has been rendered contracted as ついたち for much of Japanese history. When referring to an increment of one day, the reading ひとひ was commonplace. In fact, two days had also been expressed with ふたひ rather than ふつか, but for three days and longer, the use of the counter ～日（か） was obligatory in the same way how it is today.

※二日（ふたか → ふつか）and 二十日（はたか → はつか）are irregular. 

※As for 八日（やうか）, this is an example of ウ音便. 
※When combined with counters 十（とを） normally becomes contracted to と, but 十日（とをか） is an exception to this norm.
※The ゆ in 六日（むゆか）is thought to be a case particle equivalent to the case particle の which was already an archaism by Old Japanese that lives on in very few words. In Classical Japanese, むいか as well as むよか existed as dialectal variants. 

※As for なぬか vs. なのか for 七日, the former came first, but both existed throughout Classical Japanese with the latter believed to have been a dialectal variant which arose in Eastern Japanese dialects. 

※The use of い over いそ for 50 is indicative of more ancient language.
※The use of もも with counters appears to have been limited to 100, with ほ being used across the board for 200 onwards. 

Other colloquial names for certain days of the month also existed which don't necessarily fall in line with the native number system. For instance, the 30th could be referred to as つごもり. 

Additionally, to refer to the 11-19th, you could use the pattern "なかの + # + day counter" instead. Thus, the 15th could be referred to as なかのいつか instead of by the longer とをかあまりいつか. Similarly, the 21st-29th could be referred to as "あとの + # + day counter." Thus, the 21st could be referred to as あとのついたち instead of by the longer はつかあまりひとひ.

1. また、治承（ぢしよう）四年（よとせ）水無月（みなづき）のころ、にはかに都（みやこ）遷りはべりき。
Again, around the sixth month and fourth year of the Jishou Era, there was a sudden capital move.
From the 方丈記.

2. おほかた、この京の初めを聞けることは、嵯峨の天皇（てんわう）の御時、都と定まりにけるより後、すでに四百（しひゃく）余歳を経たり。
To begin with, as for hearing about the rise of Heiankyou, it has already been four hundred years since the capital was determined in the reign of Emperor Saga.
From the 方丈記.

3. 十一日（じふいちにち）の月もかくれなむとすれば
When the moon of the eleventh day was about to disappear
From the 伊勢物語.

4. 八雲（やくも）
Eight/a lot of clouds

5. 同じき七日（なぬかのひ）
On the same seventh day of the month
From the 平家物語.

6.
万葉仮名: 一坏乃 濁酒乎 可飲有良師
漢字かな: 一杯（ひとつき）の濁れる酒を飲むべくあるらし。
It seems better to drink a cup of cloudy sake.

From the 万葉集.

7. 二人して打たんには、侍（はべ）りなむや。
If two people hit (the dog), would it still live?
From the 枕草子.

8. 六日（むゆか）、きのふのごとし。
The sixth. The same as yesterday.
From the 土佐日記. 

9. 官（つかさ）・位（くらゐ）に思ひをかけ、主君のかげを頼むほどの人ひとは、一日（ひとひ）なりとも疾く移ろはむとはげみ、時を失なひ世に余れて期する所なきものは、愁へながら止まり居（を）り。
People who pinned their hopes on people of rank and relied on their favor, even for a day quickly try to use their energies to move, and people who were left behind by society and had nothing to hope for stay put and complain. 
From the 方丈記.