Even more distracting than the shadows would be the fixed perspective. Unless you include some form of head tracking, the immersion will never feel real. Unfortunately, it'll limit one person per room.
Using the same idea as 3D projection, you could support multiple people per room, but with reduced frame rate and the requirement that they wear glasses.
With 6 kinects in a reasonable configuration, I imagine you could achieve some decent head tracking. Probably not good enough to be convincing, though.
Classic example of head tracking: http://youtu.be/Jd3-eiid-Uw