I agree to a certain extent
Having learned my way around Photoshop (at least for the fairly basic stuff I wanted to do) I found the most offputting aspect of Gimp at the time was the UI - not least because the version I first used had something like four different windows, each with specific and not particularly obvious purposes. The fact that the menu structure was nothing like what I was expecting, and the documentation wasn't particularly good, was enough to put me off for ages. I'm clearly not the only person to think this, as there's still some crowd who basically repackage GIMP into a more Photoshop-like UI as Gimpshop, which is to my mind much more usable.
But when you get right down to it...there's remarkably little that you can do in eg Photoshop Elements that you can't do or approximate in the likes of GIMP or Paint.NET. I'm not arguing that they're up to the task of replacing Photoshop CS5 for everyone who makes a living in graphic design - but on the other hand if you put the time in and learn your way around the software you can do with a combination of TeX, GIMP, InkScape and Scribus what would previously require Photoshop, Illustrator and InDesign (or, worse, something like QuarkExpress). I say this on the basis that I know several freelance illustrators who've put together small print runs of their own comics work using said software packages.
It is interesting, though, that one of the less-appreciated aspects of commercially-provided software is the documentation and UI design. Good UI design takes a lot of time and thought, and in the context of a lot of FOSS packages which still have substantial dev input from the CLI crowd I can understand that putting a lot of effort into the less interesting (from a coder's perspective) task of UI design is not going to be a popular choice.