The grishin alignment file is just used for threading, and considers each residue in turn. I don’t believe you need to have any special chain separation designation in it — simply align up your template and target sequences in-order, as one long string.
The critical thing when modeling multiple chains is to have the appropriate chain break annotated in your fasta file. This is simply a slash (‘/’), placed between the last amion acid of one chain, and the first aa of the next. If you feed this slash-designated fasta file to both the partial_threading and the rosetta_scripts portions of the RosettaCM run, I think it should work how you want it to.
.
