Yes, this can be a bit tricky since we interpret the band gap as the region where the DOS is zero, or very small. But it typically doesn't go to exactly zero in a numerical calculations, as you can very clearly see this in a log plot.
So there is no "correct" colorscale, it depends on the magnitude of the data and what you consider small enough to be "almost zero DOS".
Your final point is very apt. It can be hard to obtain (or even define) the exact value of the barrier from just this plot, but when you compute other quantities like current, it doesn't matter of course. When analyzing the plot, the key is to keep the coloring consistent when you compare different results, so that you have a consistent way to computing changes in the barrier height etc.