At this point I would sleep on it and see if some crud was wiped from last tries (:
Anyways, If the signal is routed to the relay board and out the common RF out in one of the tests, trace the tracks out and see what the difference is. I'm hesitant but I don't think it's on the U3S board, rather something on the relay board.
Could a faulty relay be the problem ? Easy to measure... beep it out while powering the relays manually ? The LPF's are DC short so should work as jumpers for these tests.
Study the experiments that was carried out and resulted in the rev 4 relay board, perhaps some of the changes will lead you to the culprit: