I see several flaws with this test: 1) They are using data transmission rates which can be altered by outside influences very easily as their gauge of performance. I seriously can't believe they did that. It's possible to get a signal meter on the iPhone 4 if you know how. 2) The test they did (outdoors) does leave some room for inconsistent reporting - even when the phone is stationary.
It's ironic that these MIT guys talked about "amateur hour" and yet their setup was about as amateurish as you can get, under cell conditions and speed bandwidths outside of their control or measurement. Good for interest, but hardly sophisticated or in-depth. Consumer Reports used a shielded room with their own base station emulator. You can't get more consistent testing than that.