The precision that's important is the projector. You project the top left 'pixel' to the top left of the board, it retro reflects back in your general direction from the top left of the board. Same for bottom right and all the other pixels.
I presume the better the retro-reflectivity the more efficient the system is (because less light is wasted going to places that aren't your eye), but it only needs to be 90 degree ish good enough for selectivity
My son bought a pack of red-blue 3d glasses and we were looking at anaglyphic images online and enjoying them even though the images weren't perfectly clean.
In 3-d movies (Toy Story 3d) there are places where an isolated really bright object shows through cross channels and there some double imaging but it doesn't break the 3d.
These work for most people
https://en.wikipedia.org/wiki/Magic_Eye
and the channels aren't separated at all, your brain are doing it all out of desperation for closure. (Blood in the Gutter!)
That being said, circular polarization off of a retro reflective screen is one of the two state of the art ways of doing step stereo 3d: if you saw Avatar in a modern theater, there's a good chance that was used, so crosstalk should be minimal and much better than with an anaglyph.
[edit]
I should also note that circular polarization is different from the linear polarization that was used e.g. in Disney attractions in the '90s (and thus still used in some of the older attractions). With linear polarization if you tilt your head 45 degrees to one side or the other you will get an even mix of the L/R channels in each eye. Circular polarization does not have that issue.
Though it seems to me with the projector originating from the same plane as the glasses, circular polarization would be redundant. I'm sure the engineers there know more than someone who dropped out of physics after flunking the optics lab though :)