I find the most important part about reading a book to be what happens before I start. Specifically, the process I use to select what to read. We're given a very small amount of time in life, and reading a meaty book, especially non-fiction, isn't quick. So I want to make sure I'm getting the most out of the experience by filtering as many bad books as possible beforehand.
I use my book reader as a kind of reading list. When I get book recommendations from conversations or HN threads, I'll download the free sample to read later as soon as possible. This way, I never forget about a potentially interesting book.
I also actively hunt for book recommendations in threads on HN, which pop up from time to time. There was a really good one recently, but I've lost the link. They're easy to search for, though.
Usually, I don't start reading the sample right away because I'm already reading another book.
After I'm done with a book, I go through my unread samples, pick one, and begin. If the book hasn't hooked me by the end of the sample, I stop reading. Sometimes sooner if it's really bad.
The second most important part of reading a book is to discuss it as you're reading it. This usually takes the form of in-person conversations or online. I find this helps me retain the material and discover what I've missed.
You'll never retain anything about a book if you don't try to apply it in some way. This means writing, talking, and doing should all be key parts of reading. If the book doesn't lend itself to these activities, it's probably not worth reading.
Also on the subject of retention, I find reading books in a themed series to be helpful. For example:
Sapiens -> 1491 -> 1493
The first book got me thinking about the relationship between human activity and megafauna extinctions. The second book showed me that most of my assumptions about what pre-columbian America was like were almost completely wrong. The last book taught me to take the long view of globalization.
All of these books changed my mind on topics I thought I understood, which is another signal I use when selecting a book. If it can't possibly change my mind, it's not worth reading. Books that don't pass the test typically: (a) lack a well-developed point of view; or (b) approach the subject from a direction I'm already familiar with.