He's made a slight error, uncanny valley is not that as something comes to look more realistic, we notice the slight differences more, it's that as robots become visually more realistic, we notice the problems in their movement more. As they become visually realistic, we start to think of them as a real person, and then when they make movements totally unlike how people move, it unsettles us.
If there is a parallel with motion control, it would be motion control getting so good that we notice something else as a problem, not motion control getting so good that it's somehow bad.
So I think the whole premise in his argument is wrong. Uncanny valley is not going to be solved by making true photo-realistic graphics, but by making truly human like motion (down to the fine details on faces, which motion capture isn't so good at). And his noticing Natal not being perfect is just him naturally criticizing something for not being perfect. It's not like we didn't do the exact same with the much less accurate Wii remote from the beginning.
A game I'm developing with some friends:
www.xnagg.com/zombieasteroids/publish.htm
It is largely a technical exercise but feedback is appreciated.
