Because no one single alternative is vastly superior to it.
Microsoft won the Office wars in the mid 1990s by bundling all the major applications (Word, Excel, Powerpoint, later Outlook and possibly Visio, I don't recall) in a single package that cost less than buying the alternatives (WordPerfect, AmiWord, Lotus123, Quattro, etc.) individually. WordPerfect (the most popular word processing program at the time) got caught flat-footed with the shift from DOS to Windows 3.1, in part due to illegal monopoly actions as alleged in a still ongoing suit by Novell (http://www.sltrib.com/sltrib/money/52577046-79/novell-micros...).
Effectively, Microsoft poisoned the well: individual competitors couldn't outflank it, were starved for cash, and couldn't collaborate on a single alternative. Sucks to play in a proprietary world when the racetrack owner's got a horse in the race.
My own history: I'd previously used a few word processing / text editing tools, including MacWrite, WordPerfect, WordStar, vi, DOS EDIT, EDT, EVE, and AmiPro. I always found Word to be fussy, and during this period started using Linux. I recognized the value in learning a single tool that couldn't be crushed by competition or abandoned by a single vendor, and so returned to vi (as vim) for virtually all my editing, with some fill-ins from Emacs, Lyx, and the odd usage of Abiword (quick letters for postal mail) or OpenOffice/LibreOffice (if I actually have to deal with Microsoft formats). Occasionally Google Docs. And a lot of web forms (most of which go to a spawned vi editor via Vimperator, such as this one). Office97 was the first and last time I really used Microsoft's suite (though I've certainly been exposed to it since), and the less I have to be exposed to it, the better.
What I've mostly found is that text editing suits my needs far better than word processing / DTP. And there are very few compatibility issues.
And I've found that text mostly meets my needs too. But that's partly because I'm technical and willing/able to deal with formatting codes for things that should be rendered as more than text: html, restructured text, markdown, etc. In a sense it's a step backwards to WordPerfect and their split screen.
There's a lot to be said for wysiwyg. Most people are unwilling or unable to learn or use formatting code in plain text and no immediate rendering of the result.
But fundamentally there's no reason why anyone should have to deal with plain text source that produces nicely rendered output (unless we want to). Computers are really, really good at displaying things nicely, and hiding details.
The problem is that Word (and the creeping Sharepoint ecosystem) are much too powerful for almost everyone's needs. My work is typical: all our documents are a single font, and they have tables. The docs are kept in Sharepoint, and we do nothing with Sharepoint but keep documents.
That over-engineered solution would be better replaced with a very simple word processor like WordPad or some similar rtf tool, and the file system.
I don't need to use the truck that takes space shuttles from the vehicle assembly building to the launch pad just to get to work.
The key problem with WYSIWYG is that it doesn't exist any more. The key point as I read of the original article was that Word is a WYSIWYG tool aimed at print.
The Slate article could have been a lot better if the author noted the distinction between WYSIWYG and WYSIWYM. Followed immediately by the standard observation that WYSIWYM is a tougher nut. Most of the solutions I'm familiar with rely on either markup (HTML, LaTeX, Wiki parkup, markdown, etc.) or structural conventions (my text docs strongly resemble 1970s typewritten documents in line-length, paragraph style, etc. formats).
The great thing about LaTeX is that your document is entirely semantic, and the style is applied by the stylesheet. With tools such as Lyx, this is reasonably intuitive, though stylesheet production itself is a nontrivial task. Similar concepts exist for Web, obviously, with HTML and CSS.
The problem is that for the typical user, print remains a simple conceptual model to understand, regardless of how brittle it is in a world with mulitple output formats. There's been a lot of thought put into this area over decades, and for a good tool to emerge and become prominant is yet another tough nut.
I agree (from a limited and somewhat old exposure) with your assessment of Sharepoint. I've mostly seen it used as "Christmas Tree" in which individual document blobs (or other objects) are hung, rather than a Wiki in which documents are interlinked, readily refactorable, and dynamic.
