Also try testing the difference between XP's startup and Vista's unhibernation.
That would be unfair. A fair test would be xp's hibernation and Vista's hibernation.
And running the tests only with defaults on - that doesn't provide a realistic result. Especialy if the performance of non-default stuff (e.g. AV realtime) is affected by the OS.
Does XP have hibernation? I forget (it's over a year since I used the beast).
If so then by all means.
Basically the test should not be for like-like operations, but for getting your box from an overglorified paperweight to something you can do something on.