The following general sources to error are considered as the main problems:
Wrong energy equation: For example, the gravity term is not included in many commercial packages.
PVT with black-oil model and simplification of the energy equation. The pressure effect on temperature is often not included as it is normal to assume that the enthalpy change is given by Cp*dT and that the energy equation is Cp*dT = dQ = heat loss to the ambience over the section dX. For rough calculations this is often good enough approximation, but especially for wells and risers with substantial effect of gravity and pressure drop it is not good enough.
Black-oil modeling in general: To account for deviation from ideal fluids (Joule Thompson effect), flashing/evaporation and condensation is difficult with black-oil models. Normally it is however agreed that precise temperature calculations for gas and gas-condensate systems is not possible with blackoil models (should only be used for oil systems).
Compositional PVT: If not made properly compositional models may be worse than blackoil, especially near the critical point. The enthalpy for a fluid is a direct function of fluid density and it is well known that liquid densities are often very wrong with EOS models like SRK and PR. Wrong liquid densities results in wrong enthalpies for the liquid. (Since it is easier to tune liquid densities with blackoil models than with compositional models, better enthalpies and hence temperatures may be calculated with blackoil models. Very often the composition is not known and then there is nothing to gain by using compositional models).