From pure systems control point of view, your current estimation error is increased because you calculate the third phase from the combination of the the two, it’s a simple calculation. If you could live with that, no other drawbacks. You could even do this from one phase, but then the error is quite high.
Edit: This is where Kalman filtering could come in handy, because that’s exactly why the filter was designed. I’m not sure however how much work it would take. 2D or 3D Kalman should be fairly straightforward, I think. This is such a mature field, there are numerous papers on that subject.