You’ve got to also remember that any background tasks will have a impact on the results as they will interrupt the benchmark processes.
Absolutely - I turn everything off - even Airport/Wireless if applicable.
As @LightBulbFun indicated, 2.2.7 does seem best for Leopard (apart from on my MDD!) - in future I'll stick to 2.1.2 for Tiger systems as I've been using that for years.
I sometimes use my makeshift benchmark test for a more real world assessment but even that gets thrown off by using different versions of Quicktime, in fact for that test, my new MDD beats my Dual 2.3 G5, presumable because the G4 has Quicktime 7.5.5 instead of the G5's 7.6.4