Force row level locking on update statement

SQL Server 2000 SP3. Looking for a way to force row-level locking with an update statement. The table in question is very small -- has fewer than 20 rows (into 5 columns) so memory consumption is not a concern. The application uses this "lock" table as a semaphore to prevent a specific process from running while this particular process runs. A series of inserts are run on several tables as part of the transaction. After the inserts are complete, the transaction is committed and the table lock is released from the "lock" table.  The problem is that the table locking on this "lock" table prevents identical processes from doing their inserts at the same time. There are several components running the same process. The series of inserts takes about 20 minutes, which at about 3 invocations per hour is not sufficient.  Effectively, the processes wait until the table lock on the "lock" table is removed before they can run their inserts. With row-level locking on this "lock" table there would be no waiting, and the inserts could execute  roughly in parallel. Even if table-level locking occurs on the insert tables, at least some work could be done. With Oracle, we do not have this problem as row-level locking occurs.

SQL Server always uses table level locking regardless of rowlock hints, clustered indexes with fillfactor 1, and techniques such adding 100,000 new rows to the "lock" table to trick SQL Server into doing page level locking. Alas nothing has worked -- SQL Server places a table level lock during the update.

Any ideas appreciated.