get an increased energy after energy minimization when using different score functions for repacking and energy minimization

Hi, 

I am a new user of rosetta. Now I meet some problem that I cannot solve. I am wondering whether you could please help me. I would much appreciated. 

I have a protein peptide complex and want to optmize the sequence of peptide to get a better binding affinity. I am running ddg_monomer.linuxgccrelease.  I mutated single residue of my peptide in my peptide-protein complex to the other 19 amino acids. I use ddg_monomer to calculate ddG of mutations. I tried two trials, "repacking sidechain in the 8A radius of muated residue" and "repacking sidechain in the 8A radius of muated residue followed by sidechain minimization".

To do repacking sidechain in the 8A radius of muated residue

I use soft_rep_design.wts as my weight file

To do repacking sidechain in the 8A radius of muated residue followed by sidechain minimization

I use soft_rep_design.wts as my weight file and I use pre_talaris_2013_standard.wts for -ddg::minimization_scorefunction,

score12.wts_patch for -ddg::minimization_patch

I found that calculated dG of my wildtype complex is higher in "repacking sidechain in the 8A radius of muated residue followed by sidechain minimization"  than in "repacking sidechain in the 8A radius of muated residue".

I thought that I should get a lower energy if I did sidechain minimization. But it turn out the dG of wildetype is increased after energy minimization. I guess it might because I used differernt  scorefunction when I calculate ddG and do energy minimzation. 

I found in paper that they found better correlation with experiments results when soft_rep scorefunction is used for repacking and hard_rep is used for minimization. 

I am wondering whether someone can tell me which socrefunction to choose when do energy minimization. 

and whether I could use soft_rep_design.wts for both repacking and energy minimization. 

I also find three soft_rep weight files. I am wondering what are differents between them. 

I am wondering whether some experts could please answer my question.

the following are my flag file to run two trials 

flag file for just repacking  side chians within 8A of mutant residue 

-in:file:s dreria_com_2.pdb
-ddg:weight_file soft_rep_design
-ddg:iterations 50
-ddg:dump_pdbs true
#-ddg:mut_file 2ci2.mutfile
#-database /home/andrew/minirosetta_database
-ignore_unrecognized_res
-ddg:local_opt_only true
-ddg:min_cst false
-ddg:suppress_checkpointing true
-in:file:fullatom
-ddg:mean true
-ddg:min false
-ddg:sc_min_only false
#-mute all
-ddg:ramp_repulsive false
#-unmute core.optimization.LineMinimizer
-ddg:output_silent false
-ddg:opt_radius 8.0
#-score:fa_max_dis 9.0

flag file for repacking side chians within 8A of mutant residue  followed by sidechain energy minimization

-in:file:s dreria_com_2.pdb
-ddg:weight_file soft_rep_design 
-ddg:iterations 50
-ddg:dump_pdbs true
#-ddg:mut_file 2ci2.mutfile
#-database /home/andrew/minirosetta_database
-ignore_unrecognized_res
-ddg:local_opt_only true
-ddg:min_cst true
-ddg:suppress_checkpointing true
-in:file:fullatom
-ddg:mean false
-ddg:min true
-ddg:sc_min_only true
#-mute all
-ddg:ramp_repulsive false
#-ddg:ramp_repulsive true
#-unmute core.optimization.LineMinimizer
-constraints:cst_file dreria_2.cst 
-ddg:output_silent false
-ddg:opt_radius 8.0
#-score:fa_max_dis 9.0
-ddg::minimization_scorefunction  pre_talaris_2013_standard
-restore_pre_talaris_2013_behavior
-ddg::minimization_patch score12

I also extract values of dG wildtype, dG mutant and ddG from log file. and I attached values for two trials.