OK. The relevant aspect of the energies computed by CHARMM is that they should always be used in a relative sense, so you should always compute the difference wrt some reference state. This means that even though the energy values you compute are all positive, there may still be an attraction between the two rings, as long as there is a minimum in the energy surface. If I remember correctly the minimum conformation for two benzenes using the CHARMM energy function is where the edge of one interacts with the flat surface of the other (more or less).
There are also interactions within each ring, which are probably positive, but not important for the interaction between the rings. To avoid this complication you can use the interaction_energy command (energy.info).
Last edited by lennart; 03/26/21 03:59 PM. Reason: added interaction energy