John and everyone else
Thank you for your insights -- they make a lot of sense. I gave this quite a bit of thought and decided that it would be best to make an end-to-end measurement first before digging into this any further. So I added T2 back into the circuit. This way, the QDX is unmodified apart from a missing T1. I injected a -50 dBm signal into the antenna port at frequencies between 12 and 18 MHz in 200 kHz increments. That covers the same range as my 20 m test. I configured these frequencies into WSJT and used the vertical decibel indicator in WSJT to quantify the amount of RX signal. I recorded the frequency / dB pairs during a sweep. the following plot shows the terminal sweep at the left, the somewhat flawed manual VNA measurement reported before in the center (unaltered, the issues that we discussed are still there), and the new end-to-end measurement at the right.
In my opinion, the terminal sweep is still odd given the other two measurements...
Regards
Matt
