171 TYR in Chain-A of HLA-02*01 gets corrupted after second round.

Dear developers

I found that 171 TYR in Chain-A of HLA-02*01 always gets corrupted in all conformations generated after second rounds (those in directory labeled with 1 or greater).
More precisely, different carbons(C/CZ) and oxygens(O/OH) within 171 TYR show exactly the same coordinates as below:

> ATOM 1398 C TYR A 171 2.725 7.201 16.919 1.00 0.00 C
> ATOM 1406 CZ TYR A 171 2.725 7.201 16.919 1.00 0.00 C
>
> ATOM 1399 O TYR A 171 2.168 8.443 16.861 1.00 0.00 O
> ATOM 1407 OH TYR A 171 2.168 8.443 16.861 1.00 0.00 O

where CZ and OH are actually in correct coordinates, and thus, C and O are in wrong coordinates, as a results, Chain-A backbone is disconnected between 171 TYR and 172 LEU.
This is reproducible for any epitopes as long as the allele is HLA-A02*01 (I haven't confirmed if the same happens for other alleles).
Strange thing is that the conformations generated at the first round (those in directory labeled with 0) are no problem at this point.
I also checked template pdb files chosen by default (namely 1DUZ.pdb and 1i4f.pdb) but they are also fine, so I think something wrong happens after the second round.