The library is burning to the ground, and you have time to save one thing: a book on a shelf, a digital photo of the book and its pages, or or the book or manuscript digitally transcribed (that is, typed into a computer file or files). Which one would you save and why?
In imaging this scenario, I'm going to make a few assumptions. First, I'll assume that in its digitization efforts, this Library has not followed one of the core tenants of digital preservation best practices, which is to store multiple copies of digital files on different types of storage devices in geographically disperse locations. Ill advisedly, the digital files representing this book are on a single server in the basement of this burning building. Second, I'll assume that the book in question is not a mass produced paper-back of a recent New York Times best seller, nor is it a first pressing of the Gutenberg Bible. Let's imagine something in between- a not particularly rare volume, with some inherent intrinsic value: a copy of Nine Stories inscribed and signed by Salinger. Third, I will expand on the notion of a "digitally transcribed manuscript" and say that this file is a TEI mark-up of the text done by an emerging scholar in American Literature, perhaps with descriptive bibliographic annotations.
The least valuable text in this scenario is the digital photos of the book and its pages. It has neither the intrinsic value of the physical book, nor the added intellectual value of the marked up and described text. With the book in hand, new digital images can be produced. The only thing that would be "lost" with this version of the text would be the investment of resources (time and money) that went into the initial digitization efforts. The digital image would be the third priority for saving from the flames.
I would give second priority to the marked up text with bibliographic description. This file constitutes original scholarship, drawing on the expert interpretation of the researcher. TEI encoding amplifies the usefulness and reach of the text in a digital environment. Bibliographic description can be used to compare this copy to others of the same edition. These are all valued attributes. I've given it second priority for two reasons. First, given the habits and practices of present day humans engaged in the digital world- as a scholar engaged in TEI mark-up would most likely be- it is almost certain that another copy of these files exist- on the scholars personal computer, in a Google drive account, deposited in an institutional repository. Unlike this hypothetical Library, the best of us backup our hard drives, and the worst of us are too lazy to clean up our directories, resulting in a proliferation of digital copies.
Finally, the best mark-up and description of a text with any intrinsic value can never replace the original. Salinger signed the thing, and there's probably a story behind that, and people will always find the texture of that pen on that paper interesting and valuable. With the original in hand, digital copies can be produced, which then can be marked up by multiple scholars. However, there's no way to go backwards- to reproduce the value of the original from the digital texts.