1) It is reduplication of work, since it sets out to do exactly the same thing as Wayland. Supporting the Wayland process would have been better.
2) It is completely and totally owned and controlled by one company. Wayland and X as developed by many companies, and none of them has complete control.
So even if Mir turns out to be technically much better, these two arguments will still stand. Canonical's track record on other projects is a cause for some skepticism, but we'll have to wait and see.