1. All meaning is in the text. A good default stylesheet is preferred over having to design and test a custom one. (Most novels.)
2. Type and cadence has meaning and connotations that a default stylesheet can't convey. (Bringhurst's The Elements of Typographic Style.)
3. Strong and distinctive aesthetics are a part of the message. (Twyla Tharp's The Creative Habit. Constance Hale's Sin and Syntax.)
4. Design as UI, essential to the reading, but doesn't require a complex layout. (Most computer/programming books. Jesse James Garrett's The Elements of User Experience.)
5. Design is dense with meaning, fancy layout needs. (Tufte's books. David Pye's The Nature and Art of Workmanship. Josef Müller-Brockmann's Grid systems in graphic design.)
E-reader platforms serve the first type of e-book very well and, since most designers who think they're working on type 2 books are actually working on type 1 books, they serve type 2 books very well as well.
(Bringhurst's book was the only book I found on my desk where the typographic style was, appropriately, an integral part of the book's meaning. I'm sure there are others, but they aren't common. In any case, the web platform isn't capable of delivering the typographic nuance necessary for type 2 designs. Not until the spec for CSS font-feature-settings settles and is more broadly implemented.)
Type 3 books are largely let down by e-reader platforms as all of them require the ability to adjust margins and set backgrounds properly, something that all major e-reader platforms prohibit, even iBooks and Kindle Fire.
(I'm counting Apple's .ibooks format as a separate platform. It isn't a major one yet.)
Type 4 and 5 books are possible to a degree in Apple's .ibook format, but even those are partially let down by its inability to use embedded fonts. (Just try and open one up to embed fonts. Crasherama.) Those books would also, unfortunately, only work on iPads.
Some of these books (types 3, 4, and 5) are available on the Kindle platform but have had their personality and style completely removed in their e-book versions. Charging anything at all for these mutilated remains is too much. That publishers pretend that these editions are in any way equivalent to their printed versions borders on criminal fraud.
Even worse: There is no way for readers to discover which is which when they browse e-bookstores. They aren't given the tools to tell the books that have been mutilated in their e-book transition from those that haven't.
Fixed layout e-books, as implemented by Apple and Amazon, don't solve any of these design problems as they are all pre-paginated. These books are text-heavy. Doing them in fixed-layouts would be insane. It would prevent text from reflowing when resized (where resizing is even possible) and involve a massive amount of work to create a sub-par book.
IDPF's Fixed Layout spec would solve most of these problems because it adds several important capabilities that Amazon's and Apple's implementations lack:
These three capabilities mean that IDPF's Fixed Layout specification is actually useful to regular books, not just illustrated books. It would let a regular text-oriented e-book have a beautiful title page and elegant section divisions using fixed layout pages, something that many designers have been struggling with during the entirety of the e-book transition so far.
Incompatible file formats reduce innovation and harm the e-book market. The time that e-book designers spend implementing several versions of their e-books is time that isn't spent on finessing the design and implementing new, innovative, features.
It holds back the entire e-book market and sabotages the transition from print to digital.
At least during the browser wars we only had to deal with two incompatible approaches.
Fully supporting the IDPF Fixed Layout spec and letting designers both adjust page margins and set full-bleed backgrounds would cover more than 90% of the publishing industry's visual design needs.