I made a mistake when calculating the write offset for this disc, the header I got with px_d8 was for the second sector, not the first. So the offset is actually -588 samples, not 0. I know it isn't relevant for data-only discs, but it's wrong anyway.