Testing results are coming out right now. I've got the following units having at each other:

Dismounted Feudal Knights, England
stat_pri 13, 0, no, 0, 0, melee, melee_blade, piercing, sword, 25, 1
stat_pri_attr no
stat_pri_armour 7, 14, 6, metal
stat_mental 9, normal, trained, lock_morale

Dismounted Norman Knights, Sicily
stat_pri 13, 0, no, 0, 0, melee, melee_blade, piercing, sword, 25, 1
stat_pri_attr no
stat_pri_armour 7, 8, 0, metal
stat_mental 9, normal, trained, lock_morale

I'll be controlling (letting sit) the DFKs.

Pre-combat mouse-over shows 13/27/1 (atk/def/cha) as DFK stats, 13/15/1 for DNK, which is as expected. Results are DFK kills/DNK kills.

01. 59/61
02. 53/61
03. 61/43
04. 57/61
05. 61/43
06. 51/61
07. 54/61
08. 61/39
09. 45/61
10. 47/61
11. 47/61
12. 39/61
13. 49/61
14. 49/61
15. 61/58
16. 61/59
17. 45/61
18. 37/61
19. 51/61
20. 61/43
---------
1049/1139

Difference of 90 kills, 90/20 = 4.5 kills per test run advantage for the DNK, which from observation appeared to be due to getting a few extra kills when they charged into the stationary DFKs. As the result favors the unit with shield removed instead of the one with equal skill added, one would conclude that if anything the shield detracts a little more than its indicated value, but from my observation and how close these results are, I feel good saying it detracts a roughly equal amount. We'll know if my observation is correct as soon as I run off results for the vanilla units. If they come out roughly the same, I win