"David Monarchi" <david.e.monarchi@gmail.com> writes:
> I don't understand why the use of a single inequality ( > or < ) causes the
> planner to filter on the condition, but the use of an interval (> and <)
> causes the planner to use the index.
There's something awfully wacko about the rowcount estimates you're
getting. What PG version are you using? I think you may be getting
bit by the recently-fixed bugs in choose_bitmap_and() ...
regards, tom lane