R-factor of (f_obs/k, f_model) is not equivalent to R-factor of (f_obs, f_model*k) (where k is an array with non-equal elements). Therefore I believe the next snippet contains an error:

Corrected code should be something like:

        bulk_solvent.k_mask_and_k_overall_grid_search(
          f_obs.data(),
          core.f_calc.data() * k_total_,
          core.f_mask().data() * k_total_,
          k_mask_trial_range,
          selection_use)

@pafonine @olegsobolev