DOH!  I finally understood an earlier comment someone else made about <=
being worse for stable sorting.  This is because with a partial rather
than total order,

(< x y) !=== (not (<= y x)), but rather
(< x y) === (and (<= x y) (not (<= y x)))

It seems plausible that this is why CL did what it did.  Probably more
efficient...

David