It was experimentally confirmed (see the paper) that Mg diffusion length during the growth of p-layers and further annealing of the structure is about 10 nm or even more. Thus, the last nominally undoped GaN barrier adjacent to p-GaN is probably p-doped, So users need to make only modification of the last barrier doping, and this change in input data leads to about one order of magnitude increase of the predicted current.
Note it is rather common issue we have faced many times. When you making a training, it would be good idea to recommend users add p-doping in nominally undoped layers adjacent to p-layers.
