[modeller_usage] changes in MODELLER9 optimization & DNA denaturation

Hi. I just upgraded from MODELLER8v2 to 9v1 and am experiencing problems when modelling a protein-DNA complex in that the DNA is denaturing after optimization. When I use the same modelling script in 8v2, the DNA stays in its double-helical form, even when i use md_level= refine.very_slow.

Can anyone explain to me what changes in the optimizers were made that cause this denaturation? I am using the MODELLER residue types ADE, GUA, CYT, and THY for the modelling.

Thanks!

ryan