We want to know where a website actually lives, not just where its domain says it lives. We do that by measuring how quickly the site responds from different places around the world at the same time. If a site loads fast from Sydney and slow from Tokyo, odds are it sits somewhere in Australia.
The tricky part is getting measurements from inside the South Pacific. Most of these countries do not have public data centres available for us to rent a VM for doing measurements. Instead we have to rely on a combination of tests run locally while we visit the area, geo data related to DNS and IP information and triangulation tests from different locations.
We are open about the limitations, and the Pacific provides us with a unique cases where islands are far enough apart that we can see a difference in latency based on the hosting location. The goal is not to be precise down to the island and town that the service is hosted. It is about getting a solid enough picture to say: the government website you assume is local? It is actually hosted three countries away.