Never mind my previous question about using trjconv for fitting in RMSD calculations. It does produce identical results compared to using gmx rms alone. I just had a subtle error in atom numbering, which resulted in bogus RMSD values.