[modeller_usage] DOPE score based model selection

Dear Modeller_usage members,

Is it necessary to select the best protein model based on DOPE scores?
I have made 100 models of a ~260 residue protein (enzyme) using Modeller.
Then I chose the best model based on the DOPE Score and did loop
refinement on a 16-residue long loop (50 refined models created) located
at active site cleft. In loop-refined model the loop occludes the active
site cleft, but in another loop-refined model (with intermediate DOPE
score) it is alright. Can I use this loop-refined model? Also, is it
essential to refine the loop even when some initial Models have good
conformations of the loop? Any suggestions would be very helpful.

Thanks in advance,
      Vipul