fix spring/self and periodic boundary conditions

Thanks Steve,
I think I know what is going on now. For my simulation (free energy
calculation through Hamiltonian switching) I need to equilibrate my
system at a finite temperature, then I store the state using fix
store/state, then I minimize the system, attach the springs using fix
spring/self, then I restore the equilibrated configuration using a
custom restore/state fix. I use "x y z" (wrapped coordinates) in the
store state fix, which effectively destroys the data in atom->image.
If I use "xu yu zu" (unwrapped coordinates) it should give me the
right result, right?