Hi,
Since Tom Terwilliger is off-line through Jan 21 I dare to offer a guess: transferring the heavy atoms to the refined model is simply not implemented.
(...)
I think I'd keep the current protein model, but I'd discard all waters and add the Br before continuing with the refinement. I think once the Br are in the PDB file they will survive when you run the wizards.
I'm agree with Ralf. Just to add: once you have added back all Br ions don't forget to run phenix.refine again with ordered_solvent=true option (to restore removed waters), possibly with more than default number of macro-cycles (like 10; default is 3). Also, I would refine individual occupancies for all Br ions and remove those with zero occupancy values after refinement (of huge B-factors). You can do all this in one refinement run. Pavel.