This happens because of the compound angle formed between the dovetail angle and the gib's taper angle.
Only because your measurement reference is the face of each component in your model do you get a compound angle. If you take a section cut anywhere along the length inside the gib, the angle remains that of the original dovetail / gib / dovetail angle (120* in your model).
I believe
@Susquatch was pointing that out also in and earlier post.
In real life, the angle between the two parts does not change. If it did, one part would have to physically twist - which it does not.
The three mating surfaces remain co-planar.
The gib just moves up (or down) within the two constraining dovetails. Yes, the distance between the two dovetails changes (if allowed to move) based on the amount the gib moves up or down.
In the Y-axis case, an upwards movement of the gib tries to increase the distance - but of course it can’t so it jams up the saddle.
If the gib could move down (it can’t because of the way it sits on), the distance would decrease if all surfaces remained in contact.
The above two conditions assume the pivot point of the gib is at the thin end bottom - which could be reasonable since the rear adjusting screw forces that side of the gib down onto the way.