Component Data - Temperature Range
If we look closely at the enthalpy data for the formation of Fe2O3, we see that there is a phase transformation at approx 950 K from alpha to beta crystals, and another at 1050 K from beta to gamma crytals, and then melting at 1895 K. For each there is of course a change in the shape of the enthalpy vs temperature curve.
Using a polynomial approximation of a data set introduces some errors - and it is important to know if these errors and significant for your model requirements.
A cubic spline fit could be used, and this would improve accuracy - but reduce our understanding by making it easier to forget about the morphology change. The proper way to improve the fit is to create additional components for each of the phases - or morphologies that are of importance to us - each defined for the temperature range for which they apply. Then you can use the component with the best data fit for the particular conditions.