This paper just makes me incredibly sad and makes me wonder why anybody would want to deliberately put himself on the wrong side of history and write such a damaging text against LaTeX when it is already under attack from all sides; when it already requires so much effort just to convince journals to even accept LaTeX submissions. For what? To further entrench Microsoft's monopoly? To free journal copy editors from the nuissance of having to deal with submissions in other formats than Word? Or is this simply the revenge from the author for past hardships that he had to endure when working with "a Computer Science group" (mentioned in the last comment)?
As has been pointed out numerous times by previous comments to this paper, the task that has been tested here has nothing to do with the actual process of writing a paper, but rather is more akin to the final layouting phase done by journal copy editors. It probably accurately reflects their view on LaTeX and explains the resistance frequently encountered when submitting a LaTeX manuscript.
But from the point of view of an author - or a reviewer for that matter - this article and its conclusions just completely miss the point. The metrics used here, like words written per 30 minutes or number of typos and spelling errors are completely irrelevant in practice. Also the claim that LaTeX offers no way of tracking changes is just plain silly. Used together with a versioning control system like SVN and an editor like e.g. emacs it is much easier to compare to earlier versions and to compare and merge versions from different authors. Who wants these changes nicely marked in the document can use for instance "latexdiff". All of this addtional power is completely transparent and configurable.
And this brings me to my final point, where LaTeX really shows its strength, and that is its scriptability. In the task set in this paper, participants had to manualy create and populate a table with lots of numbers. This is silly and simply not the LaTeX way. I have never typed a table with more than 10 entries, simply because its tedious and mor importantly very error prone. Whenever I change my analysis slightly I would have to retype everything, thank you very much. Instead there are numerous packages that allow to directly create LaTeX tables from e.g. inside R or Python, which is not only a lot more convenient but also guarantees that the numbers given are correct and up-to-date.
I could go on writing about the power of LaTeX forever, but this comment is already getting much too long. In essence, I believe the authors of this paper have simply not understood and certainly not embraced the LaTeX mindset, which is admittedly more adapted to the needs in science and engineering and less to the needs of writers in the humanities, secretaries, or copy editors. Instead of accepting this fact, the authors go about like a school yard bully picking on the nerds and teasing them for being so different and awkward - when they are made to do things that they are simply not good at. Just let the nerds be and do their things their way, in the end we will all benefit from it.