Oh I was under the impression you had the old card still.
The difference between the 2.66GHz and the 3.0 GHz will be some, but not much.
Anyways, if you find older benches with 8800GT and a MacPro 1,1 there would most likely be with older drivers, and drivers does matter too.
If you need exact comparison, you'll have to do the tests with same drivers in the same machine.
The macbenchmark site you posted, just seems like a little blog, that has only been active for one month back in 2010.
Suffice to say, the 5870 is a giant step up from the 8800GT and just enjoy it instead of focus on exact percentages gain.
anyways:
I believe (I want to believe!

) that main bottleneck in Barefeats tests was the CPU clock (2,66 vs 3GHz) along with the sadly bad optimization of Mac games (vs Windows). This is why I would like to run the same test on the same computer with two different video cards, once o MacOS and than on Windows.
Why don't run the barefeats tests and compare against those results?
Anyways, for some reference, my hack reaches 18600 points in 3dMark06 and 11200 in 3dMark Vantage,
you should beat those by a decent margin, unless there are some bottlenecks.