Since the two methods give you readings a couple degrees C apart, it's not clear that simply averaging them will give a better number than either one alone (which one??).

My actual sensor 10K thermistor (unknown manufacturer). is Yes, I could measure three reference points to derive the three Steinhart-Hart coefficients. Or what calibration method do you mean?

Just use an accurate thermometer as a reference. Measure three or four temperatures, see what resistance your thermistor gives for them, and interpolate the rest. I'd bet the result would be better accuracy than using Steinhart-Hart on a generic thermistor. Or maybe compare to an LM35, checked at freezing and boiling -- those are really linear.

That's neat -- from the front it appears very organic and curvy, but from the sides you get a more angular look. I like it!