Parmed - the python implementation, appears to be able to write different forcefield files, such as .prmtop, and others. Therefore, rather than merging .frcmod, which suffers from certain theoretical issues, it might be more precise to load the two actual forcefield files from ligand A and B, and then combine the terms in a fully controlled manner.
This also removes the problem with "fake bonded terms" that are necessary to create the .prmtop currently. Furthermore, this relies on a python library which makes the code better - and easier to maintain.
Parmed - the python implementation, appears to be able to write different forcefield files, such as .prmtop, and others. Therefore, rather than merging .frcmod, which suffers from certain theoretical issues, it might be more precise to load the two actual forcefield files from ligand A and B, and then combine the terms in a fully controlled manner.