You can customize the CPU utilization threshold in Targets -> Hosts -> double click the host you want to customize -> select "Metric and Policy Settings" in Related Links -> double click the pencil icon at the right of "CPU Utilization (%)".
This will bring you to a screen that allows you to modify the warning and critical thresholds for the CPU % metric along with the number of consectutive times the threshold must be exceeded before an alert is triggered. For CPU Utilization (%), the defaults are that the metric is monitored every five minutes, the critical threshold is 95%, the warning threshold is 80%, and the threshold must be exceeded six times before an alert is triggered. So, if you have an alert when CPU Utilization (%) exceeds the critical threshold, CPU usage will have to measure above 95% every five minutes for thirty minutes before the alert actually gets triggered. If you want to be notified for, say, anything above 90% for 15 minutes, you would change the "critical threshold" to 90 and the "number of occurrences" to 3.