Broken structure of local docking of replicadock2

Hello, I tested local docking of AlphaRED (ReplicaDock2) protocol. As the mc steps, the fa_rep score increases rapidly.
Checking a trajectory with high fa_rep score, there are clashes between some residues. (e.g. 1st score and 2000th score)

SCORE:     score     fa_atr     fa_rep    motif_dock    atom_pair_constraint      omega    p_aa_pp    rama_prepro    I_sc    temp_level    temperature        rms     maxsub
SCORE:  -1242.286  -3616.527   2487.477      -320.722                   0.000    269.748    -98.770         36.507 689.825         1.000          0.800      0.036    627.000
SCORE:  34813.739  -3880.623  38639.354      -208.596                   0.000    278.419    -92.060         77.245 -67.071         2.000          1.500      1.113    627.000

I figured out that the backrub steps cause those unfolding or clashes.
Is this a normal production of backrub?  To prevent this, how do I change the script?
Or is there any better way to change the conformation of specific residues?

Here is the rosetta script that I used. I used n_struct=4 and n_replica=3. I am using rosetta-356 version

<ROSETTASCRIPTS>
<SCOREFXNS>
    <ScoreFunction name="score_dock_low" weights="muds_2021"/>
</SCOREFXNS>
<FILTERS>
</FILTERS>
<RESIDUE_SELECTORS>
    <BFactorSelector name="bfactor" lower_bfactor="0.0" upper_bfactor="80.0" lower="1" upper="1" min_contiguous_res="4"/>
    <StoredResidueSubset name="get_plddt" subset_name="plddt"/>
</RESIDUE_SELECTORS>
<MOVE_MAP_FACTORIES>
    <MoveMapFactory name="flexRes">
            <Backbone residue_selector="get_plddt"/>
            <Chi residue_selector="get_plddt"/>
    </MoveMapFactory>
</MOVE_MAP_FACTORIES>
<SIMPLE_METRICS>
</SIMPLE_METRICS>
<MOVERS>
    <!-- setup jumps and constraints-->
    <SwitchResidueTypeSetMover name="switch2fullatom" set="fa_standard" />
    <DockSetupMover name="setup_jump"/>
    <StoreResidueSubset name="store_subset" residue_selector="bfactor" subset_name="plddt"/>
    <RigidBodyPerturbNoCenter name="rb_mover" rot_mag="0.5" trans_mag="1"/>
    <HamiltonianExchange name="h_exchange" temp_file="hamiltonians_cen.txt"  temp_stride="1000" stats_file="tempering.stats"/>
    <DockingInitialPerturbation name="init_pert" slide="1"/>
    <TrialCounterObserver name="count" file="trial.stats"/>
    <SilentTrajectoryRecorder name="traj" score_stride="1" stride="1" cumulate_replicas="1" />
    <Backrub name="bbmover" movemap_factory="flexRes" require_mm_bend="0"/>
    <MetropolisHastings name="sampler" trials="1000" scorefxn="score_dock_low">
        <Add mover_name="h_exchange"/>
        <Add mover_name="traj"/>
        <Add mover_name="count"/>
        <Add mover_name="rb_mover"/>
        <Add mover_name="bbmover" sampling_weight="1"/>
    </MetropolisHastings>
 </MOVERS>
 <PROTOCOLS>
     <Add mover_name="store_subset"/>
     <Add mover_name="setup_jump"/>
     <Add mover_name="init_pert"/>
     <Add mover_name="sampler"/>
</PROTOCOLS>
</ROSETTASCRIPTS>

hamilitonians_cen.txt is the same as the provided file.

GRID_DIM 1
GLOBAL_PATCH atom_pair_constraint = 5
1 0.8 muds_2021 NOPATCH fa_rep *= 1.0
2 1.5 muds_2021 NOPATCH fa_rep *= 0.75
3 3.0 muds_2021 NOPATCH fa_rep *= 0.5
#ETABLE FA_STANDARD_SOFT10 fa_rep *= 1.1
 

Thanks for your assistance.