I take it this is a mean average.
I think for this kind of study you'd be better off rounding all the speeds of to the nearest 512 or 256 kbps and then using a modal average. That way it doesn't matter if there are 10 people in your sample that all have 300Mbps connections on test from virgin. It would be far more informative to say most peoples connections were 1-2Mbps at peak and 10-15 off peak.
It also presents a better metric for uswitch to get people to switch. I'd imagine that on the whole most providers are very similar when you take a mean average of speeds.