I was chatting offline about this, and sent some questions to Chris Feickert in order to confirm what Adjust+ is actually reporting. Take a look at the following picture to understand the signal phases coming off cartridge coils (picture source vinylengine.com)
The right and left coils are reversed in polarity. Why? The lathe cutter heads are reversed and L/R channels 180 degrees out of phase and then the cartridge "re-inverts" the right channel. So, referencing this, then we know the following:
1. Upward vertical movement of the stylus creates a positive signal in the L channel, negative signal in the R channel.
2. Downward vertical movement creates a negative L signal, positive R signal.
3. Horizontal rightward movement creates positive R signal, positive L signal.
4. Horizontal leftward movement creates negative R, negative L signals.
Another way to visualize cart movement and resulting L/R signals (picture source
https://www.analogplanet.com/content/crazy-little-thing-called-azimuth-part-1)
Now in the case of the test record, we have single-channel test tracks and either of the top two images apply. So if the vertical movement results in 180 degree out of phase signals and the horizontal movement creates in phase signals, then splitting the difference (i.e. 45 degree motion of the stylus) should result in 90 degree signal phase shift.
I guess, then, the "ideal" crosstalk measurement would show the signals 90 degrees out of phase, for both L->R and R->L. Further, when the phase plots cross somewhere other than at 90, and if the stylus is perpendicular to the groove at that point, then I interpret that to mean the stylus and the coils themselves are misaligned. However, other factors may still be contributing, e.g anti-skate and VTA. I'm asking Dr. Feickert to confirm all of the above.
Lastly, it is interesting to note that Richard Mak (AnalogMagik creator) does *not* seem to agree that phase alignment should trump lowest crosstalk:
You can adjust azimuth until the phase shift number is as close together as possible. However, this is not a very common occurrence and Phase shifts bear a weaker correlation and it is sometimes not repeatable unless if we incorporate a lot of averaging to stabilize the number. In our opinion, Crosstalk/%THD, measurement is usually sufficient.
From:
https://www.analogmagik.com/azimuth (A good read). He is taking a more holistic view, and emphasizing that we shouldn't focus solely on a single parameter, but rather spend the time to remeasure and minimize all distortion artifacts. I'm tempted to say don't settle for crosstalk minimization, phase alignment at some odd point. But try to adjust all parameters in order to see the effect they have on bringing crosstalk phase closer to the ideal. Then, how to prioritize phase alignment vs lowest crosstalk ... Still debating that.