Dear Colleagues,
We're encountering an issue with composite model refinement and would appreciate your advice.

Our scenario:
  1. We built two separate models based on two subtracted maps of a single complex.
  2. We have a lower-resolution map of the full complex.
  3. We merged the two models into the full complex and refined it against the full map.
The problem:
When refining the merged model against the full map using real-space refinement in Phenix, the clashscore progressively worsens.
Has anyone experienced similar issues, or can you offer suggestions on addressing this problem?

Thank you for your insights.

Best regards,
Dmitry