It's possible for instance, that the individual machine excitation settings are intentionally made different - then one might expect a different load share. I'm not sure why one would do this - there may be a reason or reasons - say for the purposes of system stability or operational efficiency ...?
The reason you change excitations on two MG sets is to smoothly transfer the load from one to another. Raise the excitation on the MG taking the load and lower the MG going offline, after the synchronizers show the frequency is close to one another. Thirty-two years ago, the synchronizing lamps would flash very slowly during MG changes.
Other than that, you try to keep them close enough to share the load.