I just use greater than, less than or equal to... [_revisionDate >= 01/20/2013] AND [_revisionDate <= 12/01/2013] for example.

Have a look at the UNK table entry with Notespeek for your field.  I suspect that the first time data was stored, the system recognised it as text and thus treats all fields of that name (no matter what the definition is on any form) as text.  This is really frustrating to track down as I found out myself a while back.

To resolve, do the following:

1.  Create a field on a form and set it as a date.  Set the name to be "CompletedDate" or whatever the name of your field is. 

2.  Create a new replica of the db and then check that replica to see if you have the same issue.  If you don't, overwrite the original db with the new replica and you are good to go.

You can also try deleting the FT index and compacting the db after doing step 1 above if you can't replace the replica.

I use the [_revisionDate] in another part of the query and use the less than greater than < > symbols and that one works.

I did as you said, because it was just a test DB it didn't have any documents in it, created field on form as date/time, removed index replicated (new replica recognized the field as a date/time), removed original replica locally and replicated back from the server to the original name, reapplied the FTIndex. the local copy sees the field as a date/time, remove original replica of the DB from the server, rename the new replica back to the original name (oh forgot, remove FTIndex from new replica before renaming it) reapply FTIndex.

Thanks it now works as expected, but a major pain in the butt and a lot of wasted (unbillable) hours.