Why are Toms still using old GPU hardware and even older benchmarks (doom 3, 3dmark 05? puhlease.. ) even for a supposedly DX10 OS?
Well the only thing that could be considered 'old' would be 3Dmk05.
D3 is the engine on which pretty much all the current OGL games are based, and F.E.A.R. and Oblivion are the two most widely tested 'stress tests' for current modern gaming.
The wide availability and 'known quantity' factor of all of them make them ideal for testing. 3Dmk06 is arguably a good addition, but still keep 05 for comparo. Something to consider 3Dmk05 gives you a tough test, but large enough numbers to expose greater variability, with 3Dmk06 the scores are lower, any differences get muted. Personally I'd rather see 3Dmk01 and 03 added to see system effects as well as just VPU.
Also If they tested something like World of Warcraft the latest version, then people would complain about ATi and nV not having the chance to ensure driver compatability/stability with the new game. There's pretty much no excuse for the titles they picked, if anything the ones they picked should be the baseline for performance.
As for the leaked drivers, reviewers shouldn't compared leaked drivers, only mention that they may exist, and then mention the impact they witnessed, but for straight up comparison, until the IHV releases the driver, then it's not something that deserves to be in a test, this goes back to the floptimizations and magic drivers fiascos of two years ago. With that history, stick with those available on nV's site. If they put them on their open Dev page, then test it, otherwise it's only worth a reference.