In checking over my pdb2cif samples, Peter Keller (who has very sharp eyes) brought a bad practice to my attention. For the sake of appearance, I had placed one token by itself and the rest in a loop, even though they all belonged to the same category. He also caught me on failing to give the ...entry_id's which are keys for some top level categories. In cleaning up my mistakes, I ran across a strange situation for PUBL_MANUSCRIPT_INCL. The example in the dictionary does the sort of things Peter noticed me doing. It puts _publ_manuscript_incl.entry_id by itself outside the loop, and it does not have a valid key. The example does give what is specified as the key (the ...entry.id), but this cannot be a valid key, since if is effectively the same for all rows. I would suggest making ...extra_item the key, since that should differ row to row. The example would then be the same, simply with the deletion of the portion outside the loop.