A few thoughts:
1. I don't how easy it will be for you to compare other setups, since I know my 9500 Pro defaults to max settings in UT2K3 benchmark, but I'm not sure if a 9000 nonpro would. (You can force max settings for comparison sake as described
here at Anandtech, but I don't know if you would want to).
2. I don't see a score difference between 640x480 and 1024x768, and the later is more common for most people to benchmark at anyway.
3. For comparisons, you need to realize that a 9000 nonpro < 9000 pro < 8500/9100 << 9500 nonpro < 9500 pro ... etc.
So I don't know if my 9500 Pro scores (XP2000+, 512MB PC2700 DDR, 7200RPM HD, Win ME) would help you at all, but here they are (at UT2k3 Max settings, but everything set to performance in ATI control panel - no AA/AF, etc.):
dm-antalus
68.315216 / 123.632050 / 351.184784 fps
Score = 123.698166
dm-asbestos
69.015236 / 146.032745 / 315.810303 fps
Score = 146.201447
dm-antalus
14.706083 / 50.316692 / 88.705956 fps
Score = 50.341717
dm-asbestos
25.249929 / 61.161533 / 98.798325 fps
Score = 61.209042
The first two are the flybys, and the second two are botmatches.
If it helps at all, I'm setting up a new ATI-built 8500LE 64MB in my son's computer next week (XP1800+, 384MB PC2100, 7200RPM HD, Win ME), and would be happy to come back here at post those scores. They should be far closer to what you would expect (although probably still slightly higher with the 8500).
EDIT: For comparison to people running AA/AF enabled, negligible effect on botmatch (0-10% reduction, depending on setting and test), but huge effect on flyby (10-40%, depending on setting and test).