As the author notes in the end, it would be really interesting to do these again on more recent models. I wonder if the no context file being cheaper still stands. But then how much does the harness influence the results. It can be frustrating trying to gauge what’s influencing what and if something suddenly starts working against you.