I've done the same tests with eldoradofm.
This is the result with no input in ST.
This is after I've calibrated, no input in ST. To achieve that, I had to set the 19KHz slider to +1/4 (measuring 0 at the bottom, or -50%) and the 57KHz slider to 0. Pilot volume to 7% and RDS volume to 4% (aren't those related to the audio playing?!?)
Now I enable the input on the ST. The FM calibration tickbox is on, at the previous calibrated values. In this case there is a drop on the stereo pilot, RDS goes up and I am getting a positive phase diff.!!
]
Finally, I disable RDS and there is still something there. FM calibration is still enabled.
So I suppose that something in the chain is picking up noise on 57khz. It amplifies the original RDS signal that ST produces. The noise of course is not in any phase relation to the pilot, it's just noise, and it's causing a phase difference between the stereo pilot and the sum/total of the 57khz signals. So it's a mess.