I dont think what you want would be possible with the hardware. Audio output os the sections would have to be polyphonic, requiring either a patch point per voice, or some sort of "abstract" patch cable, which only tells each voiceboard to establish the corresponding connection. I dont believe that the hardware had such feature.
Ok the manual states that if one of the 3 output is used THEN that particular output is NOT used in the final mix... so if I understand well you can't use it to modulate itself but maybe you can use it to modulate one of the remaining 2 SG? right?