Two bugs make a right, mostly. The very last carry should be ignored and
when adding the low and high bytes they commonly would commonly produce
a carry which was being ignored. On the test disks these two bugs
cancelled themselves out, except only for track 1, side 1, sector 0. I
had noticed it was suspicously off-by-one earlier for both test disks,
but the cause wasn't obvious.
On some old 1980s disks, with real data, it changed number of bad
sectors from 224 (out of 2464) to 5 for one disk and 81 to 3 for
another.
