Separate evasion from armor?

I saw this post at Untimately, in which Brendan lays out the basics of a system which separates the chance to evade an attack from the protection afforded by armor.  It reminded me of an idea I'd been working on a while back to accomplish a similar result, but with a few extra twists.  I eventually gave it up, because it seemed to be growing a bit too complex and because it just didn't feel like D&D to me. 

The reason I started pursuing this sort of system in the first place was because the D&D combat system does a poor job of modeling a few things that I would rather like to see in my game.  It makes no distinction between completely avoiding an attack, and being struck without being threatened with actual harm, which is important because there's a fundamental difference between attacks that seek to cause injury or death by inflicting bodily harm, and attacks such as tripping, grappling, touch-activated spells, and the like that only need to make contact rather than penetrate armor.  Secondly, it looks a little absurd when characters are fighting some huge creature like a dragon or giant, where popular fiction and common sense both tell us that nimble dodging is going to count for a lot more than a sheath of metal, but the one defense stat paradigm obstinately maintains otherwise. 

I'm providing the rudimentary bits of my system here, not as a serious suggestion but as some ideas for anyone who feels so inclined to kick around.

Instead of an Armor Class, every creature would have an Evasion Class, which would be based on its size and modified by Dexterity and magical factors.  I had first conceived of it as a descending scale like AC, though there might well be merit in using an ascending scale instead.  Shields would also apply to EC.

Every armor type would have a particular die by which it would reduce the damage inflicted on a successful hit.  Leather was a d4, chain a d6, plate 2d4.

Armor would have a coverage rating, according to which pieces were worn and thus how much of the body was covered.  A cuirass or breastplate would be good for 4 points of coverage, with an open helm adding 1 or a full helmet adding 2, with 1 additional point each for upper arms/shoulders, forearms/hands, thighs, and shins/feet. 

On a normal hit, the armor's damage reduction would be rolled, reducing the damaged caused by that amount.  On a roll that exceeded the to-hit number by the Coverage Rating or more, the attack would strike an unarmored location or a weak point in the armor, and damage would be as rolled with no reduction. 

Thus, the more complete your armor, the less likely you'd be to be struck in a poorly protected location, but the more thorough your armor, the heavier it is.  The weight of armor worn would impose a cap on effective Dexterity, so a character with a high Dex would be faced with the choice of having a good EC with his bonus, or forgoing the bonus and gaining the protection of armor instead. 

All attacks would use Dexterity adjustments to hit, and Strength adjustments to damage.  Large weapons would have a -2 penalty to hit, reflecting their slowness and the difficulty of aiming a precise strike at a gap in the opponent's armor.  Weapons made for precise strikes, like rapiers, would get a +1 bonus to hit, but no Strength bonus to damage.  Weapons designed to defeat a particular type of armor, like flanged maces against plate, would get a +2 bonus to hit against those armor types. (I like Brendan's idea of dropping a die size of the armor's damage reduction better, though.)  A character wearing heavy armor, with his Dexterity capped, would likely find it advantageous to wield a big, high damage weapon rather than a low damage precision striking one. 

I never did get all of the particulars sorted out, and I doubt I'll develop this system any more.  It seems much more complicated than it's worth to actually implement it, and my level-scaled combat mechanic already makes one's combat skill a more important part of defense with far less fiddliness, but if anything in this mess looks workable to you or sparks a new idea, I'd be interested to see where it leads.