How have you determined that the generator output grounds are at a different potential to the generator chassis, and did you determine it as AC or DC?
It's quite likely that the generator output is floating and what you're seeing, if tested with a high input impedance DVM, is a small common mode signal, often capacitively coupled between a transformer primary and secondary. What do you get if you take a current reading, AC coupled and DC coupled, from the output ground to chassis ground? (Obvious, but I'll say it for the sake of safety for the equipment, take the initial current reading current on a low (e.g. uA) setting so that you only shunt the two 'grounds' together via a high impedance.)