Generally, when preparing a PDB file for use in Rosetta, you want to use a separate chain for each different (covalent) molecule. Ideally, though, the program should give you a better diagnostic message, rather than segfaulting.
Usually when I see something like “unknown atom_name: … CA” it’s because Rosetta is trying to calculate C-alpha RMSD statistics from the structure. Since the ligand doesn’t have a alpha carbon, it triggers the message. If it’s just an informative RMSD calculation, the “error” message shouldn’t be too big a deal – it will calculate the RMSD to the residues it can and continue on. If the program is halting after the error message, however, it’s some other issue.