A factor you may wish to consider is lag time (video delay) in your system. Anything beyond 6 frames becomes visible to your audience, Consider a projector is going to introduce anywhere from 2-5 frames and it does not give you much leeway in your switcher. You can optimize that by feeding the native resolution to the projector so that it is not rescaling.
For that reason we are very particular with cameras being the same and being genlocked. If you do that the ATEM works quite well.
If you must use different camera types and resolutions and rescale in the switcher then count on 2-5 frames of delay for the switcher.