For his part, General D.H. Hill felt there could not have been 27,000 CSA infantry at Antietam... claiming his own division of 3,000 was the largest by far. The balance were artillery, and cavalry on the flanks.
Yes, but going through the estimates of his subordinates indicates:
DH Hill: DH Hill claimed 3,000 infantry, but Carman calculated 5,449 infantry and 346 artillery.
Via Clemens we get:
Ripley's brigade: 1,349 infantry into action
Rodes' brigade: 3rd, 6th and 12th Alabama carried 840 into action, excluding at least the officers of the 12th. 5th Alabama was a small regiment, but no clue about the 26th; if the 5th and 26th average the same as the other three then it's ca. 1,400.
Colquitt's brigade, from Carman's correspondence:
6th Georgia: 300+ (and one company on picket not counted)
23rd Georgia: 485 men (not officers)
27th Georgia: 400 men (not officers)
28th Georgia: 250 men or 250-275 officers and men
13th Alabama: Carman guessed 310 based on a 2nd September strength of 612
Which implies that there were ~1,780 men plus officers, for a brigade strength of ~1,900
Garland's brigade:
Carman gives the brigade 756, but it was the strongest in the division on the 2nd September (though it did then take plenty of casualties at South Mountain) and the 5th North Carolina had ~625 men at Antietam. Adding officers and assuming the other four regiments are each about half the size of the 5th NC gives an estimate of about 2,100, but I'd say we could go with 1,700 to avoid overclaiming. Effectively this is assuming the other four regiments average 250 officers and men.
GB Anderson's brigade - Carman gives 1,174 and this seems fairly solid.
Divisional total:
1349+1400+1900+1700+1174 =
7,523 infantry, and 346 gunners.
After the battle (September 22) DH Hilll's division reported 332 officers and 4,739 men, for 5,071 just after Antietam. He suffered 3,241 casualties (by his report) at both battles.
He began the campaign with 9,848 effectives (September 2) and had not fought in August.
Estimates evidently... General Longstreet wrote in "Battles and Leaders"...
But look at how Longstreet constructs the number. He takes the 2nd Manassas strength (50,000), adds Anderson (4,000) and then claims 20,000 stragglers, but he ignores DH Hill's entire column up from Richmond (ca. 25,000).
Taking the straggler number as correct but adding in DH Hill's column we would get 50,000 + 4,000 + 25,000 - 20,000, for 59,000 (not 37,000) - removing casualties from Second Bull Run is down to ca. 50,000. (The number I've used is 47,000 effectives in line at Antietam.)
Note that this number is after straggling, while the number Longstreet quotes for McClellan is before straggling, and we know that McClellan's forces straggled heavily (specifically we know that 1st Corps at least got only 2/3 of its theoretical strength into line, which accords with evidence for other corps) so actually in line of battle McClellan would have more like 56,000.
It is very much in the interests of senior Confederate leadership in the Maryland Campaign to claim enormous amounts of straggling, and specifically enormous amounts of stragglers never entered the North at all. But while this happened on some scale, the idea that over 20,000 failed to do so (over and above normal straggling that happened to every army) raises difficult questions about what the hell they all ate for two to three weeks.
By contrast, the idea that the heavy straggling only set in on the forced-marches to Sharpsburg explains this much more satisfactorily (the men were only away from their commissariat for a few days).