The rotation of -CH3 group in Non-equilibrium TI

[xiaotianzhou1982]

When I substitute a hydrogen of one methyl group by -CN which will form new interaction with the pocket of protein.
In NPT run of state A, the methyl group rotate freely, but in NPT run of state B the interaction of -CN with the pocket will limit the rotation of the -CH2(CN) group. So when I do the transition from state A to state B, in which the -H rapidly changes to -CN during the 50ps Non-equilibrium TI, 3 independant runs will give different results due to the freely rotated -CH3 group. Because the initial structure of the 50ps Non-equilibrium TI is obtained from the previous NPT runs, the position of H-to-CN will be random. In some runs, the H-to-CN happens far from the interacted residue of the pocket, and in other runs, the H-to-CN happens at the interacted residue of the pocket, thus leads to different ddG. Does Non-equilibrium TI not work for such case due to short transition process? Or any restrain protocol need to be added?
Thanks
Fred