The reason why you still get this alert is the way SQL Server detects these situations. What the alert is doing is it looks in the sysperfinfo table. If you look into that table you will see that the value for deadlocks/sec = 1. Once anther deadlock occurs the value will be increased to 2. So as BOL says: In most cases, this will be a level or monotonically increasing counter that counts occurrences of the instance event. Not the actual current events.

I once ran into the same problem with replication conflicts/sec and was able to find a workaround. In your case you might build some alert which checks if the value in sysperfinfo has increased since the last check.

Hope this helps

M