Ok you guys nailed it. Sync polarity was the key - the operation of switch #4 was not properly documented in the silk screen (it specified it as being set to 'OFF' for all modes (CGA, MDA, EGA))

If I turn switch 4 'ON' in 200 lines mode it causes GBS-C to detect the source as a generic mode and works although improperly centering the image left to right. In 200 lines mode the source is properly detected as standard NTSC timed RGB with the switch 4 set to 'OFF'.

However in 350 lines mode GBS-C does not detect a source at all with switch 4 'OFF'. Turning it 'ON' in 350 lines mode the GBS-Control status monitor now shows the mode detect correctly identifying the source as medium-res and I do get an image. It's still not properly scaled, but I think Rama (the author of the GBS-Control custom firmware) can probably correct this.