The key here is the importance of contextual cues in the formation of
episodic memories. Our memories depend on tying up with the actual data we want to remember with huge amounts of seemingly-irrelevant information, which act as 'hooks' to integrate the data with everything else.
When you have a physical object, it's easier to provide the context needed to create a map. As mentioned, the place within a book is a very powerful cue, which is not surprising as our episodic memory is primarily evolved to remember places. But there are ways of translating this into a form usable for ebooks.
The first requirement is a rich and varied design - there's a reason good textbooks have large page sizes, lots of figures, sidebars, and various bits of typographic dressing, and these need to be retained in an ebook instead of squashing them down to create a linear flow. The second is good, and fast, spatially-directed, navigation within the book. This is the primary reason current low-power devices suck for textbooks - the CPUs take too long to render the page, and even the faster eInk displays take too long to refresh. Page turns need to happen in milliseconds in order to facilitate the formation of a broad overview of a particular chapter.
Navigational aids need to be rich and contain a variety of dressing and cues. This is where ebooks can improve on the physical - a table of contents is always important, but in physical books it's stuck away at the front and isolated from the material. eBooks can integrate the ToC with the material itself to aid mapping and should also allow the formation of flexible links. The devices also need to be high-resolution - every time you have to enlarge a picture or figure to see it clearly you're removing it from the context in which it was placed and you're destroying a vital part of its function.
eInk devices excel for fiction, which the narrative flow within the text itself creates a journey which doesn't need additional reinforcement. But technical works and especially textbooks have greater demands. I'm not surprised that the Kindle DX didn't meet requirements - it's just too slow. The iPad has a better chance, but it's still too low-res and a bit too small. Obviously a good device for textbooks needs to support rich annotations and marginalia, and needs to allow the user to create links within the text on-the-fly. We just aren't there yet, either in terms of hardware or the software framework - ePub 3.0 still lacks a proper page model, but it's still miles ahead of .mobi.