The only “problem” i see is that you set a flexible downtime. This means the downtime really starts if the host goes down. If you downtime is longer than 1 hour in your defined time frame than you get an alarm.
The description for the duration from the API documentation is also clear here.
duration integer($int32)
default: 0
example: 3600
Duration in seconds. When set, the downtime does not begin automatically at a nominated time, but when a real problem status appears for the host. Consequencely, the start_time/end_time is only the time window in which the scheduled downtime can begin.