It's an incomplete list, so let's not pretend otherwise.

This should make it much more obvious (via check-near-bragg) when the peak positions are wrong.

I'm happy that there are no gremlins here, now.  This happens only for 000,
where rlow and rhigh both go to zero.  It doesn't make sense to have a special
check for the condition.

I'm sure I've fixed this bug many times before...

This fixes a not-so-subtle bug introduced by 289c8a77.