not many heatsinks are designed to have the air "sucked" across them. the alpha pal 8045 is supposed to suck, but i seem to get better results with the fan pointing towards the cpu. also, i like the secondary cooling effect of the air blowing across the mobo. my brother has an slk800 with a tornado, and he has a ton of air the blows right across the north bridge heatsink, so that helps alot with high fsb setups as well.
your best bet is probably to have the fan blowing down, but test it if you want to. maybe it will work better.
In theory should pushing the air work better than sucking the air. Cuz when the fan sucks it sucks the air close to the mobo and the "hot" components. That would draw warm air over the fins of the hsf. 1 degr increase???
However when the fan sucks, the deadspot problem would be smaller... Hmmm. Not sure what I think anymore. Try it! If it works you will be happy. If it not you have at least learnt something new...
Blowing air into the heatsink is usually (but NOT always!) best.
Almost all heatsinks are designed for air to flow in this pattern although there have been exceptions in the past, and may be a few available now (I am out of touch with the newest aircooling stuff.)
Experimentation is good though- occasionally you may find a system where pulling air out of the sink will work better, regardless of what the manufacturer intended! Air flow is weird and can cause some strange things to happen- try it and see.
Reading this again I though of another point to add:
Checking temps at full cpu load is the best way to be sure which way works best.
Getting a better temp when the cpu is idle is not very relevant; getting 1 or 2 c lower at full load is (at least when overclocking to the extreme end!)