We are trying to refine a complex that was solved using Phaser. We have run a round of rigid body refinement using Refmac and now want to run a simulated annealing run in Phenix. Visually the complex looks pretty. The resolution we currently are at is 3.6 A. When we go to run phenix.refine the program quits with a Number of non-bonded interactions < 0.0001 error. Is there a way to loosen this up a bit or get around it or should we go about this in a different way?