Hi!

I believe there is an error in bulk_solvent::k_mask_and_k_overall_grid_search procedure.

1. k_overall should be calculated over selection (i.e. "work set"), not whole set of reflexes

should be

     FloatType k_overall = scale(f_obs, f_model.const_ref(), selection);

2.k_overall_best does not correspond to r_best if no better R-factor found in the following grid search. Like I mentioned above selection should be used in this case too.

should be

   FloatType k_overall_best = scale(f_obs, f_calc, selection);
   FloatType r_best = r_factor(f_obs, f_calc, selection, k_overall_best);

Note: Probably it would be a good idea to remove selection function argument (i.e. require users to pass already pre-selected arrays)