i applaud any attempt at objective testing rather than gut feel approaches to learning what matters - it's a great way to improve your skills. in this case there are a few too many variables to say what the results mean from an iso point of view.
i've done a similar test with a 40D, and there is next to no difference what iso setting you use, except that ISO200 is just starting to look worse than higher isos:
http://www.philhart.com/canon_test
for an equivalent test, the sub exposures should be the same length for each iso, otherwise you have different amounts of read noise in each and are not comparing the effect of only ISO anymore. (having said that, with an uncooled camera, thermal noise dominates and read noise becomes insignificant even with 1 minute subs).
i'm happy for people to believe whatever they want, but my testing convinces me that there is very little difference between ISO settings on a 40D but that the lowest ISO settings are not the best.
my results have no bearing on any other camera (except a 450D with same sensor). with a 20D (or 350D) previous tests were more strongly in favour of ISO 800/1600.
Phil