I just did some tests and I'm sorry to say, I'm absolutely certain your conclusion is incorrect.

here's the test...

1 unit of pike militia (player) vs 1 unit of peasant archers (computer)

normal unit sizes. grassy plain, 2 sets of 10 tests. archer's ammo capacity reduced to 10. pike militia taken out of spear wall and defensive stance. clear weather for all tests.

Set 1, unmodified pike militia with 3 armor upgrades. (armor stats says 3)

Men Remaining
test 1= 57, test 2= 50, test 3=52, test 4= 50, test 5=50, test 6=53, test 7=46, test 8=54, test 9=56, test 10=51.

average men remaining= 51.9

set 2, modified pike militia with armor stat set to 8, no armor upgrades. (armor stat says 8)

test 1=55, test 2=55, test 3=61, test 4=57, test 5=56, test 6=54, test 7=54, test 8=55, test 9=57, test 10=63

average men remaining= 56.7

a difference of 4.8

considering that the armor stat difference is only 5, I'd say thats fairly significant and isn't likely the result of a statistical anomaly.

conclusion= the armor upgrading system not only appears screw up, its also functionally screwed up. i wish it weren't so, but it is.