Accurate measurement of exhaust gas temperature in internal combustion engines is a challenging task. The most common, and also the most practical, method of measurement is to insert a physical probe, for example a thermocouple or platinum resistance thermometer (PRT), directly into the exhaust flow. Historically, consideration of the measurement errors induced by this arrangement have focused on the effects of radiation and the loss of temporal resolution naturally associated with a probe of finite thermal inertia operating within a pulsating flow with a time varying heat input. However, a recent numerical and experimental study has shown that conduction errors may also have a significant effect on the measured exhaust gas temperature, with errors approaching ~80?K depending on engine operating conditions. In this work, the authors introduce a new temperature compensation method that can correct for the combined radiation, conduction and dynamic response errors introduced during the measurement and thereby reconstruct the "true" crank-angle resolved exhaust gas temperature to an estimated accuracy of ±1.5?%. The significance of this result is demonstrated by consideration of a first law energy balance on an engine. It is shown that the exhaust gas enthalpy term is underestimated by 15-18?% when calculated using conventional time-averaged data as opposed to using the mass-average exhaust enthalpy that is obtained by combining the reconstructed temperature data with crank angle resolved exhaust flow rates predicted by a well-validated 1-D simulation.