To fully pass the test you need to get 100/100, and for the animation to be smooth and the generated image to have the proper appearance. It is a little confusing that the test doesn't automatically check the other two things.
FWIW, my Firefox 6.0.2 gets 100/100 but the resulting image does NOT look, pixel-for-pixel, identical to the reference image. Ah, but I guess it's only required for default settings, so perhaps that is the problem - non-default settings on my end.
I guess it can't test rendering conformance since that may require taking a shot of a part of the screen, which might just raise some security considerations if JS was allowing that.