This one is really easy to fix. In section 5.2.4 (Computing
Signatures), the paragraph ordering implies incorrect things about a
user ID certification signature (which hashes the public key plus user
ID packet). The description of a user ID certification signature
refers to "the data above", which given the paragraph ordering, is how
to hash a signature for signing, and not a public key.
If we just switch the position of the paragraph beginning "When a
signature is made over a signature packet" with the paragraph
beginning "A certification signature (type 0x10 through 0x13)" the
problem goes away.