I repeated the first test, but let the hot plate run longer. It revealed that the linearity for the lower temperatures completely falls apart at the higher temperatures. I think it should be fairly straightforward to modify the code to accommodate this.