SICP is not about FP, it's about all forms of programming. It starts with FP because that's the easiest model to reason about. Then it shows you how if you add state to FP, that can get you to OO. Then it shows you how you can use that to build an evaluator for your favorite programming language, or a virtual machine that can serve as a target for many languages. As a deep dive into a single topic it's not very good, but as an overview of much of computer science it's unparallelled. Knuth appears to have gone for the deep dive approach, and good on him -- but that's why TAOCP is yet unfinished.