I agree you can get better scores with different drivers, but at the cost quality/stability in other games. They noted in the artical as why they choose those drivers, they couldn't run some of the other benchmarks with the betas.
Personally I always choise drivers that have stability, quality and then speed in that order, so based on that you can hardly use a set of drivers that will only run well in a singe game / benchmark as the bases for an entire benchmark?? You do have to take into account that the drivers might be taking advantage of a 'shortcut' which will not work in the long run and the final driver might be just as slow as the release ( the 6035 betas allready have reduced the lead over the release drivers ).
Also in all the driver comparisions of I have seen only VillageMark and 3DMark 2k1 benifit to any degree from the new drivers. Games like Quake 3 seem to stay stable, and in fact have a large performance cut with the new 6032 beta drivers.
http://www.epigamer.com/article.php?a=28 (last page for 6032 update)
I will say though it is very hard to find an artical wich is trully bananced, I have seen many sites benchmarking ATI on slow release drivers to beta nvidia and vice versa (the 23.11 are stable, but not very fast in comparision to 27.xx).
Practically all articals that I have seen recently (that don't appear to have a huge bias) seem to put the 8500 between the Ti200 and the Ti500, and given John Carmack (id software) comments, particularly regarding the high polygon scenes (ATI think there is a may be a HW flaw, ie not fixable by a new driver) I have trouble seeing why many people consider the 8500 to be faster than the Ti500.
Nvidia needs to convinced it's manufactuers to life their game when it comes to the RF filter and improve 2d quality, ATI defentily has a lead there. 3d is more debatable, particularly since since SmoothVision is simply a modified version of SuperSampling that Nvidia has since their GF2 cards but since replaced with faster routines since the performance cut was WAY to high. Nvidia and 3dfx where in a simular conflict when Nvidia was comparing SS against 3dfxs AA, 3Dfx refered to SS as a 'driver hack' back then and I can't see why SmoothVision can't be considered in the same light.
http://www.anandtech.com/video/showdoc.html?i=1562