Is stockfish goning crazy?

I solved the puzzle of the day and went into analysis of the whole game and when I analized it it showed castling as an inaccuracy and the move suggested as better was castling. Did this happen to anyone else?

I'm guessing that if that's reproducible, it's a communication error about the two different ways castling can be expressed to the engine (for example, I think SF recognizes both e1g1 and e1h1 as castling moves)

Thank you :)