You do not need to define separate trajectories. Instead, you can rerun to recalculate energies for the relevant energy groups. These can be defined using an index (ndx) file, although this may be not even needed, depending on your topology. If you include the desired sets of energy-grps in the mdp file, generate a new tpr file and then run gmx mdrun -rerun with the new tpr file, you should be able to get what you want in the energy (edr) file.
A note of warning however from a previous question.
And also, I am assuming your groups are covalently bonded, so non-bonded interactions are all there is.
I hope this helps,
David