I understand what substitution-permutation networks look like. I understand what Feistel networks and Feistel-like networks look like. A Feistel network or a Feistel-like network features substitution and permutations, so why is it considered distinct from an SP-network?
In the substitution step of a typical Substitution-Permutation Network (e.g. in AES SubBytes), the whole state is broken in parts and each part substituted. That's not the case in (the core of) a Feistel cipher, where at each step/round some sizable part of the state is bound to remain unchanged (in order that each step be reversible).