Oops, that's actually right!
The correct hash for part 77 actually is matching the one listed in the project itself. Apologies for the confusion, and the source of error is exactly as you described. Why it did happen and "corrected itself further down the line" - because I used an external tool to which I ctrl+a and ctrl+v 'ed the content of a <pre> tag. Which omitted the trailing line break (which haven't happen before nor after). When I run SHA256 algo on the full content directly, it generated "2f654...".
Once again, apologies for the confusion.