A graph of the percentage error for each of the calibration data points for this particular Chebyshev fit is shown  in Fig. \ref{432198} below.  Notice that although the fractional error in resistance is largest at the lowest temperatures, the fractional error in temperature is actually largest at high temperature (because of the lower sensitivity).  The Jupyter notebook containing the Python code used to calculate T from R and to generate this figure is attached to Fig. \ref{249499} above.