*Second reading*: Finished Chapter 8,9,11,12 and also Chapter 2 for my comprehensive exams. I still hold onto my earlier view that this book is really a reference text rather than learning stuff from the very basic, although it is also true that I think that there is unlikely any textbook that can go simpler than this one (though they can be less comprehensive but spend more effort on something else). One way to tell whether a book is reference text or not is to see if a beginner finds the book size overwhelming.
For the most part, what I have found very useful for studying with this book is to complement it with other books or lecture notes. The reason is very simple: quantum information is actually very close to pure mathematics in their arguments, although mostly linear algebra. Linear algebra is easy to grasp but hard to master. For this reason, even some proofs in the book can appear very demanding and unnecessarily complicated without the right tools. Two examples I like the most
to illustrate this is (1) the proof of strong subadditivity theorem, and (2) the proof that any CPTP map has the operator sum representation. For (1), the issue is that if Nielsen/Chuang introduced monotonicity of relative entropy first, then the proof comes very quickly, but they decided to do it the opposite. For (2), if you learn from other books/lecture notes about "Choi representation" and vectorization map, then the proof is actually very natural: you are trying to "invert" Choi matrix of a quantum channel by constructing relevant states to take partial trace. But that's simply inverse-vectorizing the eigenvectors of the Choi matrix. Hence the proof looked opaque for beginners (like me).
There is also the proof about fidelity as maximization over all purification; the proof used two purifications that are not at all obvious unless you know the answer already; in costrast, if you look at Wilde's lecture notes/book, he simply used "canonical" purification that anyone can write down for any density matrix in five seconds (this is also done in Watrous' textbook, using vectorization language). The proofs are all equivalent in essence, but the presentation is more confusing and unnecessarily general.
Overall, a nice book to keep and refer to, but you are better off having something else with you while keeping this one in your bookshelves. I recommend keeping this together with Mark Wilde's and John Watrous' textbooks; you will probably need not to get any other texts beyond these.
--------------------------------------------------------------------------
Stopped at the place where I used for my QI course. This book is more like a reference text which you should keep with you at all times for referring to details or refresher, so it is somewhat unsuitable to read cover-to-cover unless you are very free.