Applicability of different temperature oscillation calorimetry algorithms, previously proposed in the literature, for the evaluation of the heat transfer coefficient during the course of the reaction, is analysed by referring to the experimental data from a methyl methacrylate polymerisation reactor. Results are coincident in the nominal case (no errors on data), while differences appear in the elaboration of real data. The effect of uncertainty in the experimental data has been simulated in terms of error on the amplitude and on the phase of reactor and jacket temperature profiles. This approach allows to indicate algorithms having the best robustness properties and to give an explanation of phase errors in terms of a lumped parameter which accounts for different phenomena, not included in the basic hypotheses of the method. Based on this observation, a procedure to eliminate this source of errors from experimental data is proposed for the more general algorithm presented to handle oscillating temperature profiles generated by different techniques.