I haven't done a deep dive into the phenomenon, but I think it's an artifact of data collection.

I believe the question about food insecurity is something like, "In the last 12 months, were there times when you worried whether your food would run out before you got money to buy more" or something like that.

So, if you at one or two times worried about the money running out, but you were pigging out on junk food at all those other times, you'd still be measured as "food insecure."

Related comment from myself regarding sociological data: #1221367