Could you share any reference to those statements? Until now I thought that the only difference between GTX and RTX in multi-view rendering was the number of views it could do simultaneously, the former only two, the latter four, but no other restrictions. If, as you say, GTX support is limited only to co-planar projection planes then it could significantly sway my decision about which card to get.
The whole discussion also spins another interesting question. It seems that originally Pimax had to, apart from pre-lens warp, also transform the rendered results from co-planar geometry to its native geometry.
This sounds totally reasonable, except, how is it supposed to work on non RTX card? Is the engine rendering into non-coplanar planes and skipping the GTX stereo-view feature? I guess the HMD has to pass its geometry to the game engine and the "Compatible with parallel projection" simply make Pimax HMD fake its geometry as using co-planar projection planes, but at the cost of additional transformation.
This explanation however does not explain, how using (or to be exact, not using) this feature can result in 35% performance improvement, unless their transformation algorithm was really eating up so much.