Sorry your message was stuck waiting for approval because of the attachment size. I've now increased the limit to 1000 kB.
I think I have solved about 80% of your problem. Since several residues were represented only by their "P" atoms, MMB was probably not able to match structure. I have deleted these incomplete residues. I also added a model of chain "X" as an extended chain (I can do this by doing a stage 1 MMB run). I then used "loadSequencesFromPdb," which reads sequences and residue numbers directly from last.1.pdb, and tolerates missing residues (as long as the remaining residues have all their backbone atoms).
I adjusted the "threading" command to not thread the missing residues. I fixed chain A to the Ground frame (rootMobilizer A Weld). Lastly, I turned on the Amber99 potential for chain X only (with setDefaultMDParameters and "includeAllNonBondAtomsInResidues X 57 91").
So now, the only part not threaded is the span from residue 71 to 81. If you want to thread to the "P" atoms from those residues in chain A, I suggest creating eleven "springToGround" commands, connecting the P atoms on chain X to the locations of the P atoms on chain X. Look at the reference guide for the springToGround syntax. You may want a nice strong spring constant, e.g. 300 or more.
Let me know how you do with the last part. I have attached my commands.dat and last.1.pdb.
Sam