In testing car mileage, two questions arise: whether the test regimen is realistic and whether it produces results that enable you to compare one car with another. I believe the answer to the first is "no" and to the second is "yes."
The article describes a test procedure that enables the public to compare all tested vehicles. In order to be able to compare car A with car B the procedures have to be identical, which means they cannot be done on the highway, with whatever fuel is handy, by mom and pop drivers.
What the article really wants to complain about is that the test procedures are not reflective of normal driving by normal people, and produce unrealistically high mph numbers. I believe that to be a valid complaint. The test procedures appear to have been designed by the auto manufacturers, who for obvious reasons prefer the sticker mph numbers be high.
So, can you believe the sticker numbers? No. You probably won't get mileage as high as those numbers. On the other hand, can you use them to compare one car with another? Yes. The mph numbers may be too high but should be roughly the same amount of "too high" for each car. Therefore, are the sticker mph numbers useful? Yes, as long as you understand how to use them.
