Fort Lauderdale, FL
Tests do not do justice
These tests seem very biased towards CDMA2000/ev. Also, their charts show cities within their test states, but no where are they actually showing data on "283 total testing sites". So I really cannot believe they did a very thorough test in that regard. They also missed out on at&t's strongest state 3g connectivity and coverage location (Florida).
Reliability in percentages just doesn't seem acceptable. You cannot quantify having either a good connection, decent, or no connection within certain areas of connectivity with reliability percentages. What could that actually mean to an actual user of said network? That standing in some special spot, 60% of the time you will be connected GOOD, or maybe 20% of the time it will have magically NO CONNECTION?
Also, reliability means its there, ready, when you need it, at any given time. They should have done tests many different days consistently, then take an average of those days. Either way, saying "good", "excellent", or "poor" connectivity would have been a better choice.
I as a user of both att and Sprint can safely say they are very consistent wherever I travel (usually all of east coast). I usually hit over 2mbps with att here in south florida, sprint around 1.5. Now att obviously has a terrible 3g coverage map compared to the cdma guys, but aside from coverage, PC World, imo, did not do much justice with these tests, specifically with "reliability". Speeds, on the other hand, involve too many day-to-day network factors -specifically user volume/congestion on each tower.
If it ain't broke, don't fix it. Unless you know what you're doing.