This is awesome work, sorry I haven't been able to join in the conversation the last few days. Something I want to point out, you may already know this but there are two possibilities for the poles on a cylindrical magnet. The magnets I bought from digikey have poles like this:
This is important because if you put this magnet in the center of a shaft and rotate it, there is not any change of the magnetic field relative to the sensor. I know because I made this exact mistake.
One other important thing I learned is that the output of the sensor is affected by the rotation of the magnet as well as the position of the magnet relative to the sensor. This becomes important if the sensor is not exactly lined up with the center of the shaft (It won't be) because the rotation of the magnet gets compounded with the change in position relative to the sensor, and the measurements get crappy. What I found was that I got the cleanest measurements when I intentionally offset the magnet by a large amount, and lined up the poles so the rotation of the magnet would not affect the sensor, only the position of the magnet.
I also did a little concept redesign of the arms based on the design you have been working on. I think a two piece arm that clamps over the shaft would be good. I am worried about the set screw arrangement you have slipping, depending on the material you use for the arms.