With the new self-test.ini, this is what I see, with Rx1_L, Rx1_H, and Tx1_1 uFl->SMA cables and antennas all still connected:
For apples-vs-apples comparison, removing only the Rx1_H antenna uFl connector:
(Edit: Just realized that I had removed and replaced the cables since yesterday's test; the cables were the same, but the cable positioning was probably slightly different -- also, the machine had been powered off and on.) After removing all three uFl connectors from the board, so no antenna connections are attached:
The influence from Rx1_H is by far the most noticeable. Just for fun, I started poking around on the board with my finger (old microwave trick -- I was also touching the metal case). With everything re-connected, there is at least a 5dB difference in the space between the curves when I poke around the USB and power supply end, but this was the most interesting picture:
My finger was on top of the Rx2_H and Rx2_L transformers. I have no idea what this proves, other than the isolation in loop-back mode isn't perfect. Any other ideas?
(Edit: I also just discovered that the position of the box fan blowing on the board made a huge difference, and could also be a factor in my measurements today vs. yesterday being different. I had moved the fan out of the way to mess with antenna connectors, but when I put it back close to the connector end of the board the FFT result changed from basically the first picture above to basically the last. It's just a fan! No idea........)