The three attached screenshots are from the same game, on hard mode. Three different battles which happen to be at the same location. Some of the lists continue beyond what the screenshots show, but whatever is unshown cannot contradict the described errors.
"August" :
Top line shows no losses for attacker, and 384/390 killed for defender.
Detail view for Kills (Total, incl Main Phase and Pursuit) shows 4 losses for attacker levies, contradicting top line.
Detail view shows 149 Losses for defender levies, yet defender Levy survivors shown as 0/384.
"October" :
Top line shows 638 losses for attacker, and 39747/39753 killed for defender (6 survivors).
Detail view for Kills (Total, incl Main Phase and Pursuit) for defender shows at least 582+390+50+45+23 kills (total of green numbers is 1090+) yet top line says only 638 losses for attacker.
Detail view for losses for attacker is also over 1000 (41+58+961+5), also contradicting top line.
Attacker losses in red do not match the count of survivors:
Kandayat: 2769/2800 survivors (31 missing) yet shows 41 losses
Paiks: 2750/2800 survivors (50 missing) yet shows 58 losses
Levies: 14199/14731 survivors (532 missing) yet shows 961 losses
Armored Horsemen: 678/678 survivors (0 missing) yet shows 5 losses
"November" :
Top line shows 20 losses for attacker, and 6310/6312 killed for defender (2 survivors).
Detail view for Kills (Main Phase) for defender shows 50+41+2 kills (total of green numbers is 93) yet top line says only 20 losses for attacker.
Detail view for losses for attacker is (3+4+82=89), also contradicting top line.
Again the count of survivors doesn't match the numbers Killed/Losses.
We can note here that on the November screenshot, we can see the winner army on the map view does have 21772 troops, which suggests that 20 losses is correct, and 93 or 89 losses is wrong.
The White Numbers in the Detail view (count of survivors) seem to be a close, if not perfect, match to the top summary line.
The Red and Green numbers in the Detail view (kills/losses) are wildly wrong, generally overstating the effectiveness of the losers and understating the effectiveness of the winners.
I wonder whether somehow these red/green numbers are calculated prior to inclusion of damage modifiers for countering, battle advantage, terrain, etc.
This would be in line with the notable unreliability of the "green flag" / "brown minus" predictor of who will win a forthcoming battle, where again the prediction seems to ignore the post-Adventurers importance of Battle Advantage.
