That's a good way to put it all together. I always struggle with mounting bearings!
The Allegro A1321 was discontinued, which may be the one that the 1302 is being confused with here. NA85 on the ubi forums mentioned that he had contacted Allegro, and said that they were expecting to have a replacement for the 1321 out at some point. The reason that the 1321 was desirable was its sensitivity, which is 5mV/G.
This leads me to another point. In the video the shaft is rotated almost 180 degrees to get from minimum to maximum voltage. I'm not sure what your application is Gene, but this sensitivity probably isn't the most suitable for a joystick, for two reasons. Firstly, the amount of rotation required (the average stick rotates between 20 and 45 degrees I think), and secondly the output will be sinusoidal (half of a sine wave anyway), rather than linear. Even if you use levers or gears to amplify the rotation of a stick's axis, you'll still have sinusoidal output.
This might suit your particular application, but for people looking to put the sensor in a joystick (Troll) I can suggest a few things.
A linear output is desirable for a joystick, as it outputs just like a potentiometer equipped stick, and you can then apply whatever sensitivity curve you need in-game. To get this, you need to get the sensor to read from minimum to maximum voltage over an angle of 40 degrees or less. Doing this ensures that the sensor is operating only in the linear section of the sine wave (well, pretty close to linear).
There are two ways of doing this - use a higher sensitivity sensor, and increase the sensor's magnetic flux. To increase the sensor's magnetic flux, you can use stronger magnets (higher N number), use larger magnets (thicker rather than wider I think), and place the magnets closer to the sensor.
When you're trying for a linear response, it's better to initially have too much sensitivity, because you can then decrease it by making some adjustments to the geometry of the sensor (eg space the magnets wider, or change the position of the sensor so it's not directly between the magnets), or even use only one magnet.
Anyone looking to replace an existing joystick's pot would need to measure the angle that each axis of the stick rotates over, and optimise the sensor for this angle. Doing this ensures that you get the full resolution of whatever analog to digital converter that you're using, rather than needing to 'zoom in' on the output curve, or put up with sinusoidal output, for something which is typically linear. Typically in flight sims people will either leave the sensitivity curve alone (linear) or decrease the sensitivity around the center of stick movement - which is exactly the opposite of what happens when your stick has part sinusoidal sensitivity.