Duplicate rows on import

When doing an import with constraints=n, has anyone ever had the import create multiple duplicate records?

I did a test import using a single table that has 27 million rows. In production, there is a column used as a PK which thus has a unique constraint. First pass through I did an import with the indexes=n, constraints=n. import completes successfully, rows are in.

Then, I started the import again with rows=n, indexes=y, constraints=y. Import failed, inability to create unique index on the column indicated due to duplicate key values.

Huh. Tried to manually create the index. Same failure. Huh again.

I built a quick non-unique index on the column. Couple of sql queries later and I determine that there are 20,189 duplicate values. I run the same queries on my production system (where I got the export) and no duplicates are there.

I did the export of the data while the system was up live and users active on it.

Based on this, it almost seems like either my import corrupted somehow, or my export. How do I determine if my file is bad, or my import is bad? My only thought is to try and run it again or to possible see if the unique data is at the end of the table (reflecting the activity going on in the production system at the time of the export.).

--
http://www.freelists.org/webpage/oracle-l