Using VFP5.

I have a table that uses pessimistic table buffering (and is filtered). Before the user saves the data, I want to check for certain things like duplicate address, etc. I use a Select statment to do this:

select address, count(*) from gal ;
	where fbatch=galtrack.batch and !deleted() ;
	group by address having count(*) > 1 ;
	into cursor checkdup

What I have found is that the SQL statement is actually looking at the data on disk, not what's in the buffer. Because If there is a duplicate address and you fix it, the SQL statement still reports that there is a duplicate. Am I loosing it, or is this a known problem? If this is a known problem, anyone got an alternative idea?

Thanks.