Here is an example that perhaps illustrates the point better - comparing projections of 1ATP (crystal structure, molmapped at 6Å) to 2D classes of EMPIAR-10252 (same molecule, cryo dataset):
The comparison with the projections is very helpful, visually, and allows identification of some classes that might be otherwise discarded. But clearly some of the pairings are bogus, so it would be helpful to be able to manually edit/interact with the outputs.