Not as you phrased it, but I think the point still stands. No, there's not too much drawing involved in the production of (most) conversions, certainly not when compared to original tilesets. However, it happens not infrequently that you do have to make some edits of individual tiles, which will in part familiarize you with (emulating) a drawing style potentially rather more complicated than your own. More importantly, though, making a tileset conversion has the potential of familiarizing you with the process of tileset creation abstracted away from the actual drawing, and so things like tile arrangement, masking, which event tiles to include, what users will find helpful in a tileset, etc., all of which carry over perfectly to more traditional tileset creation.

Okay, in CR's defense, he said not one word about uploading the levels after so editing them. His explicit comparison was to people who redraw manga pages, who I'm guessing don't then turn around and try to sell those specifically redrawn pages as original art, especially not to manga producers.

Semi is also an extreme example because it makes very specific use of certain parts of masking that I agree are not found in many other tilesets, but I doubt the same is equally true for every other level out there. If it were, to take a similarly extreme example, we wouldn't see so many Battle1 conversions. More importantly, though, I would think that even discovering that fact about Semi, that it has those diagonal pipes and that ladder with holes in the rungs on the left and everything, would itself be a useful discovery if one's looking for ideas for level creation. The ultimate goal as I understand CR's post/s is not the editing of any particular level but the in-depth study of different parts of levels brought about through the editing process. In thinking about how something would work if changed in manner x, one must necessarily think about how it works at all, what its contribution to the whole is, and so on.