I think a close study of the Graham Phantom will explain this geometric issue better. Sorry for being so clumsy at explaining in writing.
Notice the bearing angle to the azimuth adjuster on the "magnaglide". This angle matches the angle at the headshell so when you change the azimuth of the pivot, the azimuth of the cartridge changes with it. (I know the Phantom is a quasi-unipivot so it requires outside support to avoid azimuth rocking)
Top view:
http://ic2.pbase.com/o4/06/530506/1/53878804.GraBal3.jpg
Side vide:
http://www.graham-engineering.com/db3/00271/graham-engineering.com/_uimages/Phantom20II20Magnet.jpg
_____