What one has to remember is that the "standard" picture of a p-n junction or similar is the average of several (very many, in fact) atoms, and what goes on microscopically (nanoscopically, we say perhaps) - not least in the interface between the different regions - is completely ignored. In ATK, conversely, we are concerned with what precisely goes on on the length-scale of individual atoms, in this picture concepts such as "the bottom of the conduction band as a function on z" is not a particularly useful concept.
Now, the LDOS plot does let you come a bit closer to this picture, and what you do see is that the conduction band is largely aligned throughout the structure, with almost all offset taking place in the valence band.
Without having any real data at hand, just the plot, I would assign the top of the valence band to the envelope of the blue peaks sticking up out of the white, and the top of the conduction band as the edge between the broad solid blue region and the next color value. I might be wrong, but tuning the parameters (in particular the max) should allow for some further insight. Also, it's a bit odd that you have zero density of states in such a large negative energy interval, did you really compute the LDOS down to -3 eV, or just -1? Seems you need to go lower.
