I see in 3DMark03's help file that it is "required" you set all driver settings to maximum quality to produce a score regarded as "default". Hence review sites should really be maximising this slider if they're doing this benchmark. But this doesn't apply to other tests like UT2003. And even here at Beyond3D, from the
9600 Pro review:
For the purposes of this review the default driver settings will be used for normal rendering, as this produces standard Trilinear Filtering as represented above. 16x 'Quality' filtering will be used for more detailed game results on the Radeon 9600 PRO.
So that's not in line with 3DMark's "requirements".