It's regretable that so many of these reports are missing. However, inspection reports as collected in M935, Inspection Reports and Related Records Received by the Inspection Branch in the Confederate Adjutant and Inspector General's Office, provide a useful guide for the commands represented. For instance, this report for Hiram Granberry's Brigade was made Aug. 25, 1864. Inspection cover page summary states the following --

Aggregate Present and Absent -- 2,523

Aggregate Effective for the Field -- 909

Number of Guns -- 784

I've created a spreadsheet for this report and numbers seem quite reasonable. For instance, here's the breakdown for the 10th Texas Regiment, ten companies --

Present for Duty -- 14 officers, 146 enlisted men, 160 total.

Present on Extra Duty, Sick and in Arrest -- 2 officers, 43 enlisted men, 205 aggregate present.

Absent on Detached Duty, POW, Sick, Furlough and Without Leave -- 12 officers, 103 enlisted men, 320 aggregate present and absent.

Individual columns are available for each category which I've subtotaled above.

This report describes one Confederate command as it existed towards the end of the Atlanta Campaign. I don't have inspection reports for Govan's or Polk's Brigades on hand, but the division summary for Aug. 31st shows 498 officers and 4,443 enlisted men present of 12,795 aggregate on roll. Considering this number includes five brigades (Olmstead's Georgia Brigade was assigned to Cleburne earlier in August), these are seemingly reasonable division totals for this period of the war.

Here's my point -- a report for the 10th Texas Regiment made during October of 1862 should list 750 to 1,000 officers and men on roll, about forty percent of the total being present. Twenty such regiments at that time should report fifteen to twenty thousand on roll, about six thousand present and four thousand actual PFD. Of course the number of effectives depends on how many arms in good repair are available.

Earlier posts in this thread suggest that Confederate reports from the company level and up don't accurately represent commands on the field. If actual evidence to demonstrate this situation exists, I would like to know more about it.