Researchers who use statistics needs to understand statistics. It can be a useful tool, but it can trick you, and it is often used to trick the audience. This article is a good example of bad stat.
It looks like they make very poor comparisons. Worker diet vs. retiree diet does not account for a lot of other variables. If they had compared single male workers to single male retirees, they might have got a good comparison. As it is, they found that when a man's wife dies, his diet goes to pot. Gee. Big surprise. A lot of guys can't cook. Not me, not all you foodies out there, but a lot.
Maybe compare the 64 year old diet to the 68 year old diet. Are those really different? Or are you measuring 70s diet vs. 30s diet?