Energy dependence on the position of the system in the box

[ilya]

Hello!

I am running QM/MM calculation of nonperiodic system. I noticed that
the energy of the system is highly dependent on the position of the
system in the simulation box. For example even small translations can
change the energy for 10 kcal/mol. The system has a charged qm and MM
parts. I am using MT poisson solver for MM and QM electrostatics. The
box (both for QM and MM) is approximately twice as large as the system
as required for MT solver. The dependence is strange because MT should
be translational invariant.
I need to find the energy of reag, ts and products of the reaction.
During the optimization the geometry of the system changes and the
resulting relative energies depend on weather i center the system
every step in the QM box or not for example.
Could someone give me a hint how to make the simulation correctly. Is
such dependence a normal behavior?

Best regards.
Ilya.