Sorry - I did speak complete rubbish earlier : ) It's the Steinhart-Hart which is ultra-accurate (hundredths of a degree or smaller). Personally I'd trust that over the Molex chart! : ) But again, I could be speaking more rubbish : )
Depending on the requirement, it could be pre-calculated if there's sufficient Flash, i.e. just store the value every tenth of a degree or something, and then interpolate between them. I'm sure you know all that anyway, just for the benefit of others.
I don't know if any information there is useful. I recall quite a few bits of ADC resolution are saved if the resistor (if it's in a potential divider circuit) is sized just right for a desired limited range, and my maths certainly isn't good enough to calculate such a thing, so I just let Matlab run a few different attempts.
