First, have you analyzed your network traffic to know just how much broadcast (and multicast) traffic you have now? There needs to be some broadcast traffic on a network for content delivery and network discovery. You need to know how much you have now, and what type of broadcast traffic you have. That will also give you a baseline to measure your control method and its effectiveness.
Then, have you analyzed the APs where isolation is not working? What's different (brand, model, connected devices, type of traffic, is there traffic difference with isolation enabled and disabled?) I'd look there first. Secondly, go up the network from there, any difference in switch or router type or configuration? Compare your answers with the AP that is performing isolation and discover the differences.
As for methods: yes, port isolation works -- on APs and any access switch ports going to your APs. On any services switch do not enable isolation. The exact steps will vary depending on vendor, but the admin console should give you an enable isolation option. Isolation does not prevent broadcast traffic, but it does stop those frames from going to the athX interface and then being transmitted. (see: https://help.ubnt.com/hc/en-us/articles/115001529267-UniFi-Managing-Broadcast-Traffic for details -- it has a good description of broadcast traffic.)
You can listen on the switch's wireless interface and broadcast traffic will not be received by computers connected to an AP. Future network analysis down to the device level (although do not use tcpdump, you won't see the blocked traffic as blocked) should show a decrease in broadcast traffic reaching end users.