Sure. a) It's necessary to have no alpha-channel for the layer otherwise the erased parts in c will gett transparent and you would have to re color it with white which makes the borders (which you get by shrinking the selection by 1 or 2px in c) look unclean (at least to me xD)
b) This step is for things like the clouds which you can see on the left example picture. The red line is my addition on the second layer. Now you could ask "Why can't I draw it just on the original picture?" I believe it's easier this way because you cann correct much more easier than drawing it into the picture. If you made a mistake you would need to be careful not to erase the original lineart and that's why I suggest to first draw it into a second layer. When finished you can merge both.
c) Select black areas, shrink the selection by some pixels (so you get a border-line) and then start erasing.