It's pretty well understood that government tests are not reflective of real world driving habits or conditions. The test is however consistent and can be used as a basis for comparison of different vehicles.
It's very much like synthetic benchmarks for CPUs. They won't give you a 100% accurate picture but they're a good basis for comparison.
I would wager almost everyone skews numbers at least part of the time. This is one of the main reasons we have regulation in the first place. They simply can't be trusted to do the right thing.
Look at the fuel/distance range some are advertising with. I never once got the same fuel/range ratio or the difference is a lot bigger in practice.