Hi Hongmin, One should not be seeking to arbitrarily reduce the Rfree value, which is an important indicator of agreement between the model and the data. I believe the discussion of the 'multiple copy refinement' approach in the context of the ABC transporter concerned the fact that it helped to disguise the significant errors within the structure. In effect, you are increasing the number of parameters to be fit several-fold by refining several copies of your structure, decreasing the observation-to-parameter ratio, which of course is already low in the case of a low resolution structure. It is probably better to look into possible reasons for the high Rfactor/Rfree values...
Cheers, James
