Laplacian photoluminescence-based local diode voltage evaluation was shown recently to lead to correct mean values of the local saturation current density in mm-sized regions, but local maxima in the positions of recombination-active grain boundaries appear overestimated. It is shown here by 2-D device simulations that this effect is at least partly due to the influence of the local diode back voltage, which is caused by the voltage drop at the bulk and back contact resistances. Visually the image of this back voltage appears like a blurred copy of the local diode current density. It is shown in this work that indeed the diode back voltage can be simulated in good approximation by blurring the diode current image, which comes out of the Laplacian evaluation, multiplied with an effective vertical bulk resistance. The corresponding point spread function can be obtained e.g. by device simulation. An iterative procedure is proposed leading to self-consistent results for the diode current density and the diode back voltage. If this method is applied to simulated local cell data, the assumed distribution of the saturation current density is retrieved accurately. Applying this method to measured photoluminescence images leads to a better correspondence to non-linear Fuyuki PL evaluation results than the previously performed direct evaluation of the local diode voltage data. Remaining differences will be discussed.