Hi, we had the same problem, and we solved it adding the -use_pdbseq keyword. We believe this is connected to a numbering problem. I.e., in the pdb file for the docking the residues numbering doesn’t start from 1. Please, let me know if this solves your problem.
Real thanks for your help to our post. We were able to build libraries as you suggested, and to start some loop rebuild (though we have some other problems now). Let’s try to cooperate on this, since no help comes from the developers (sigh!).
> Hello,
> I have been having a world of trouble getting this exact scenario to work as well. Required is a .fasta file with the run name (t000.fasta) and fragment libraries (which I generated from Robetta and renamed aat000A03_05.200_v1_3 and aat000A09_05.200_v1_3). I also have a pose_loops file called t000.pose_loops containing a single line with my start and end residue defining the loop that I would like modeled. And finally there is a t000.pdb file that contains protein 1 (chain A) and protein 2 (chain 
separated by a TER statement.
>
> When I run the following with the above files, I see the subsequent error:
>
> rosetta aa t000 A -s t000.pdb -dock -pose -loop -randomize 1 -nstruct 1
>
> “Number of residues in pdb file 420 disagrees with total_residue: 314”
>
> So I inferred that my .pdb file containing both proteins should contain only one protein. Sure enough, when I remove the second protein and run the same command, I get:
>
> “Docking partners are goofy….this simulation’s hosed. Did you forget a TER statement?
>
> Any help is weclomed