It's pretty simple really. The EPA measures tail pipe emissions and backs that number into how much fuel was burned. They put the car on a dyno that varies resistance in both start/stop(city) and long haul(highway). Since in low acceleration city driving, hybrids are running on battery, they produce, exactly 0 emissions and it skews the average way way up. Furthermore, hybrids come into the test with fully charged batteries, which consequently doesn't account for the gasoline that ultimately charged the batteries in the first place, further skewing the results.
Essentially, the parameters are known for the EPA tests and automakers came up with a beautiful and completely legal way to beat the test.
Fascinating. Makes perfect sense. Thanks for the explanation.