Hi @fernandesj,
Have you already tried Non-uniform refinement? It’s specifically designed to optimally treat micelles in membrane proteins and does not require masking or subtraction, often achieving substantially improved map quality and resolution for protein domains. Check out the preprint here for results and details:
In our hands, non-uniform refinement generally works better than subtracting the micelle (since the micelle is actually present in every particle image, it’s just disordered).