Correlation is not causation. While it would certainly be nice if true, and it certainly passes the common sense test, this study does little to prove the causation.
For example, it could be that people who have the necessary genes for maintaining their better physical health are therefore more likely to stay active in their old age. So it could be their genetically better health leading to the behavior, not the behavior leading to the better health.
so what you're saying is that people who exercise are genetically predisposed to exercise, and those who don't, aren't. therefore any study that suggests the health benefits of exercising should take this into account.
Twin studies (http://en.wikipedia.org/wiki/Twin_study) aim to examine the differences due to genetics vs environment. One naive method is to look at identical twins who have been separated, and see if there is a difference in exercise levels.
For example, it could be that people who have the necessary genes for maintaining their better physical health are therefore more likely to stay active in their old age. So it could be their genetically better health leading to the behavior, not the behavior leading to the better health.
I wish science reporting handled this better.