Hi Matt
There are two properties in play, feature operation and table handling.
If you are always replacing the table data entirely then insert feature operation and drop and create table handling is simple. This will also create the table if it doesn't exist.
If the table has relationship classes though, dropping the table will also drop the relates (which Data Interoperability cannot recreate), in which case you need to insert features with a truncate existing table handling.
Most elegant is to also read the existing state of the target table, use a ChangeDetector transformer to figure out the delta, then use fme_db_operation feature operation to do insert, update and delete operations as required. Be aware though that ChangeDetector is brutally strict, things like date precision matter if you include datetimes in the compare.