I'm bad at math (and I joined a bitponzi scheme to prove it) but the idea is that the larger the pool, the less variance will occur because it solves more blocks in a given length of time.
BTCguild was unlucky last night, not to mention that it was the middle of some growing pains.
Give it another shot or three and you'll have a more valid comparison.
So mathmatically speaking, do you make less by joining smaller pools?
A smaller pool will (be likely to) pay out less frequently over a given length of time, but total payouts from both will converge over longer time periods.
Over time, that extra 3% you're paying deepshit will add up.