If you want Rosetta to consider arbitrary ligands, you need to supply a ligand parameter file so that Rosetta understands what its atoms are and how they connect. This is true even if you want the ligand left alone. Check the ligand docking documentation for how to generate these parameter files (look for “molfile_to_params.py”.
Including the parameter file will hopefully allow relax to handle the existence of the ligand.
That said, it’s possible that docking will still refuse to recognize your ligand (it won’t crash, but it might ignore it). If the ligand is not at your interface, you’ll be okay; if the ligands are at the interface this might or might not cause problems. I suspect that the docking code will try to dock the first two chains in the PDB together and ignore the other chains, I’m not sure.
If docking does cause problems, you’ll have to go into the docking code and tweak its foldtree a bit – this is not hard but it’s scary if you don’t know C++. Basically you’d need to ensure that the rigid-body jump between the two proteins was modified, but that the rigid-body transformations between ligands and their proteins was not.
If you want to do docking between ligands AND proteins, you’d be better off waiting for the multi-ligand docking code to get released (hopefully soon).