Now, they're mostly pretty inexperienced players. With some of my previous player groups I'd likely only kill 1 Fighter, maybe none if the dice favoured them.
LIke you said in an earlier post, this is the difference.
If your players (and by extension their characters) don't understand the basic rules of the game with regard to tactics and how to use them, you are going to have a much, much different experience. This says nothing, however, about the weaknesses or strengths of the CR/EL system, though.
The tactics involved in a troll-on-fighters fight, BTW, are very basic. (They have to be, because there's no magic, no rogue abilities, no nothin' but hittin' stuff and movin' around.) If a player doesn't understand why it's (often, but obviously situationally) better to provoke an AoO from a troll than let it swing at you with two claws and a bite, that player is operating his fighter
at least a level or two below the fighter's actual level.
Noumenon, the reason a seemingly small AC boost makes such a big difference is due to the multiplicative aspects of probability.
Assume a creature with two claw attacks. Both claw attacks do respectable damage, but the
real damage is if both claw attacks hit. Assume that normally the creature has a 50/50 chance at hitting a fighter with each claw. The chance for it to land its serious damage is thus 0.5 x 0.5 or 0.25. 1 time in 4, it will clobber the fighter.
But if the fighter can bump his AC just a little, to the point where the monster only hits on 1 claw attack in 4, the chance of it landing its
big damage goes down drastically, to 0.25 x 0.25, or 1 time in 16!
So each claw attack probability was only halved, but the rend probability was quartered! That's a game changer. And the basic principle holds for normal attacks, just to a lesser extent.