It's an odd book, and one I actively recommend against when I'm mentoring new developers. My objection is similar to yours: while there are some decent ideas in there, there are also some howlers that are simply Martin's personal preference and that are not supported by (or even that are actively contradicted by) more objective and broadly drawn evidence. No-one within the target audience of novice programmers is going to know enough to tell the difference reliably, which means the book could easily do more harm than good to someone in that position.