Dear Shunsuke, you indicated at the first line of you mail that you are working at a low resolution
I am refining a low resolution data (4.1 A).
where the bulk solvent contribution is essential ; at lower resolutions it becomes even dominant.
Start R-work = 0.3762, R-free = 0.4043 (no bulk solvent and anisotropic scale) Final R-work = 0.5019, R-free = 0.5204 (no bulk solvent and anisotropic scale)
Start R-work = 0.2665, R-free = 0.3195 Final R-work = 0.2578, R-free = 0.3126
Therefore it is normal that ignoring this component gives you a model with a very high R. So the real question is to be sure that the bulk solvent parameters have correct values (you did not give them). Looking at your total R-factors, I would believe it is OK. The R-factors perfectly correspond to the linear approximations : R = 0.091 ln(d) + 0.134 = 0.262 (you have 0.258) Rfree-R = 0.024 ln(d) + 0.020 = 0.054 (you have 0.055) (see Urzhumtsev, Afonine, Adams, 2009, Acta Cryst D65, 1283-1291) Best regards, Sacha