It sounds like you have an open circuit or a short in the temperature circuit (GPI3). The first thing I would check is that there is a jumper in place for that circuit on the 25x2 header.
It's possible. I certainly haven't tested in that configuration, so I couldn't say for sure. But the code can do some strange things if the shift inputs aren't connected (this is by design - the code must have good inputs to function properly).
If this was mine, I would try removing the jumper, and testing the signal to the CPU by using a voltmeter on the GPI3 side of those pins on the header to see if the voltage varies as you change the pot resistance. If the values do change appropriately, then I would assume all will be well once more circuits are hooked up.