Glad you're getting concerned about benching results. I'm flogging a dead horse, but they haven't been 'right' since the 6xxx series came out.
One to look at in todays review is the 8xAA here
. I'd go out on a limb and say that the reference 580 can-not be faster than either an oc'd 580 or indeed this card. (I'll bow to anyone who can explain how the same tech running slower can produce a faster result however)
I've noticed it in something like the 6970 reviews onwards. The base cards in many graphs being faster than the 'technically' faster card. This, not only is "wrong", but shows that something is wrong with the system being used for testing.
I'll harp on about it again, but I *think* it's your Windows 7 setups. Of course I can be wrong, but that's where I'd be looking.
Good write-up tho.