There are more tests to pass yet. It currently passes all but 22 of the full test suite. Mostly voodoo with the “undocumented” X and Y flags on bit 3 and 5 of the F register. Thankfully the behaviour has been documented since.
I understand that Joffa liked to use R (the top bit doesn’t get incremented automatically?) as a flag, I think he used it to track whether the frame interrupt had fired yet. Pretty cool! I know some loaders balk if R isn’t the value they expect though! Was that the issue?