I don’t think it’s possible with stock code. Rosetta is not performing a sequence alignment before the structural alignment; it assumes it is comparing copies of the same structure. If the pose lengths don’t match, it doesn’t know what atoms to ignore in RMSD calculation.
The simple option is to edit your native pose so that it will sequence-align with your result poses; whether or not you can do this depends on the problem.
The complex solution will be to rewire the code to take an alignment file between your pose and your native pose and align based on that.
I think one of the other executables may let you do RMSD calculation with explicit alignment files easily, so you could postcalculate RMSDs? It looks like super_aln has this functionality, but the RMSD part is commented out, so I don’t know if it works or not.
.
