I think you only have a chance to increase check_interval and snmp_timeouts to reduce your false alarms. Because if you have a 60s check_interval and you have to wait 70s for response it the next check is scheduled.
Do all your switches have this problems or only larger ones?