>>>Well if it was saved last night AND it was NOT the subject to a REPLACE during this process, then I am at a loss.
>>>I guess I am at a loss anyway, because IF it was saved last night then one would expect SQL to have given you the copy of the old record from disk EVEN IF you had done a REPLACE on the one in RAM.
>>
>>IAC, I restarted the application and it provided the same. So, this is also why I posted this message. I doesn't make sense that I have to GO TOP before the SQL in order to get that record from the UNION. So, for now, I put a comment on top of it explaining why I had to put GO TOP.
>
>Just to check Jim's idea see, what GetNextModified and GetFldState functions would return prior to do your select. May be there are uncommited changes?

No buffering is in effect in that application as this is a robot which does backend process at specific interval.