Parallel heat flux calculations at the JET divertor have been based on the assumption that all incoming heat is due to the projection of the heat flux parallel to the magnetic line, q, plus a constant background. This simplification led to inconsistencies during the analysis of a series of dedicated tungsten melting experiments performed in 2013, for which infrared (IR) thermography surface measurements could not be recreated through simulations unless the parallel heat flux was reduced by 80% for L-mode and 60% for H-mode. We give an explanation for these differences using a new IR inverse analysis code, a set of geometrical corrections, and most importantly an additional term for the divertor heat flux accounting for non-parallel effects such as cross-field transport, recycled neutrals or charge exchange. This component has been evaluated comparing four different geometries with impinging angles varying from 2 to 90 degrees. Its magnitude corresponds to 1.2%-1.9% of q(parallel to), but because it is not affected by the magnetic projection, it accounts for up to 20%-30% of the tile surface heat flux. The geometrical corrections imply a further reduction of 24% of the measured heat flux. In addition, the application of the new inverse code increases the accuracy of the tile heat flux calculation, eliminating any previous discrepancy. The parallel heat flux computed with this new model is actually much lower than previously deduced by inverse analysis of IR temperatures-40% for L-mode and 50% for H-mode-while being independent of the geometry on which it is measured. This main result confirms the validity of the optical projection as long as a non-constant and non-parallel component is considered. For a given total heating power, the model predicts over 10% reduction of the maximum tile surface heat flux compared to strict optical modelling, as well as a 30% reduced sensitivity to manufacturing and assembling tolerances. These conclusions, along with the improvement in the predictability of the divertor thermal behaviour, are critical for JET future DT operations, and are also directly applicable to the design of the ITER divertor monoblocks.