This is difficult. Running the jobs for the individual structures (ghost_1, ghost_2, monomer_1,monomer_2,...) each on a seperate machine would not be very efficient since they will have very different run times.
To use parallelization together with jobbsse you can just set the PARA_ARCH environment variable to either MPI or SMP and run the programms in parallel mode as it also done for jobex. This should be more efficient.