Somewhat strange methodology is used there: "speeds can be altered by mobile networks' use of proxies, meaning that the speeds recorded will be between those p

Somewhat strange methodology is used there: "speeds can be altered by mobile networks' use of proxies, meaning that the speeds recorded will be between those proxies and Akamai's servers, rather than between the servers and the actual mobile devices themselves." For users end mile usually matters the most and this table doesn't count for that, so it doesn't reflect the actual user experience. Sounds more like PR stuff than an actual test.