LOAD has the responsability to get the data in there as fast as possible.
So duplicate key are not immediately checked and are therefore put into the
page.  In the second phase of the load, the indexes are either maintained
and or rebuilt.  At that point the load will recognize duplicate keys on
unique indexes.  The row being already in the table, it is deleted.
If you run the same load with the  following ... FOR EXCEPTION
tabschema.tabname
then load will keep the first key encountered and insert into the exception
table any duplicate subsequently met.
Please note that only for duplicate keys, both the delete and insert will be
logged.

So, as previously mentioned your 89 rows were most probably duplicate keys
deleted.

HTH,   Pierre.