If the "Embedded Linux vendor" is unable to contribute the code changes back to the originating Open Source project, that should be a warning signal, either they are too incompetent (not necessarily technically but communication wise) or client's best interests are not in their best interests.
About the book, I think it's very nice overview of all the embedded stuff upto command line level. It might have had some small sections on real-time alternatives (real-time linux kernel, linux on top of real-time kernel, linux and real-time kernel side by side, either on top of interrupt scheduler or separate CPUs) and internationalization (C-library locale data can be huge, how to add locales, handling gettext and message catalogs).
Some more mentions of what could be done on host machine would have been nice too, e.g. tell about using valgrind to find memory errors and leaks before going to target.