At the risk of everybody thinking I'm a freak, I have a hypothesis to try.

As Poyer lists the last of serial numbers in December 1895 to be 24919 and stating that the Model 1892 ended that month, I believe that 24919 would be the maximum number of Model 1892's built. Again, if anybody has a better substantiated number I'd like to know it. There was mention of some Model 1896's having been built in 12/95, also. This would of course decrease the number of 1892's having been built.

The Senate report of fiscal year ending 6/30/1897 shows 3,508 Model 1892's having been altered to 1896. If Poyer was again correct that only "new, unissued" rifles were altered, then I would suppose that these would be some of the last 1892's having been built. Applying some mathematics again:
    24,919 - 3,508 = 21,411
This gets it close to my serial number 21,015 (a difference of 396). As my rifle does  have the resemblance of the alterations, I am thinking my rifle is part of the 3,508. Dick Hosmer's gun serial #20,197 was not altered and built 6-10 days before mine (120 rifles per day during this time period). So to end my hypothesis, I believe that somewhere between his serial number and mine would be the beginning of the Model 1892's grabbed for the alteration in the fiscal year 7/1/1896 to 6/30/1897.