Down duration when using 'must exist for' gives DownTime from alert start time not actual downtime

We get lots of false positives if there's a blip in network, or when a server is rebooted.  We have the polling rate at 1 minute.

We've just change one of our alerts to have the "Condition must exist for more than 3 minutes", and once a server has been down for that length of time, it correctly alerts, as it should.

But if the server is down for 3 minutes and comes back up just after the alert was sent, the reset email says it was down for 0 minutes, when actually, Orion knows it was down for between 3-4 minutes.

Does the ${N=Alerting;M=DownTime} variable clock start the moment the alert triggers? - I guess so.  Is there a way to get the actual duration the server was down for, not the duration the alert has been active?