A lot of really good suggestions in this thread
For the sake of completeness, I'll mention an approach I saw in the RQ rules mailing list, that I cannot at the moment properly credit. I have NOT tried this in a live game, so, YMMV, but it kind of intrigued me.
The approach is to roll an additional 20 sided die to track the level of success - call it a 'crit' d20.
If your main attack is a success, and your 'crit' roll is a 1, you've critical'd
If your main attack is a success, and your 'crit' roll is a 2, 3, or 4 - you've special'd.
If your main attack is a miss, and your 'crit' roll is a 1, you've fumbled.
If you have over 100 skill, you have to play some extra games to see if you roll your crit dice more than once (ie. at 125% skill, if you roll under 25% you'd roll your 'crit' dice twice and take the best result - at least, at first glance that seems to make the math work )
The math geek in me digs it because it's a restatement of the idea "When you hit, 20% of the time you will special, 5% of the time you will critical.", just moved into the time domain (wipes nose on sleeve)
It's not the most particularly elegant approach, but it does have some side benefits - it 'fixes' the low-skill range oddity where a person with 5% skill in something will 'crit' 20% of the time they actually succeed. And it makes it easy to expand the number of crit categories, if you're into that sort of thing ('special fumble' becomes easy for example). If I were to use it in a game, I'd probably do something like : to-hit roll of '01' and a 'crit' roll of '1' would get you the original RQ crit (2x weapon damage, ignoring armor) that I always found to be a bit too lethal to occur 5% of the time.
Anyway, food for thought.