Yes, this is a good idea to simplify the calculation, provided it gives a good fit to the actual curve.
Maybe it's not a good idea to adhere to the pure theoretical data, because this might not take into account some additional effects being responsible for the unlinearity, maybe some unlinearity in the thyristor driver or transformer? Why not taking your measured data (originally taken in laboratory) as theoretical data and puting this into the table? Maybe deviations are smaller if you take from this table the trend later?

You should additionally have some protection against calibration errors. Could you store into the serial EEPROM more than one set of calibration points and take the average when corretcing the curve? You could also periodically put a new set of calibration points into the serial EEPROM, while dropping the oldest set. By taking the average of these sets of calibration points you have both, a protection against deserters and a calibration which includes long term drift of machine.

Kai