People who don't do up their seatbelt buckles die.
Therefore, seat belts fail to protect people.
If researchers show people on average say they do up their seat belts but are actually fibbing, lying, or mistaken, and don't actually do up their seatbelts properly, then, yes, it does show seat belts fail to protect people.
But more importantly: it says, you can't trust people when they say they do up a seatbealt.
Which is where you and your analogy fall flat on the face.
It doesn't physically effect other people (much), if you fail to put on your seatbelt, and you lie about it -- you're hurting yourself.
If nobody follows the scientific method, or implements appropriate statistical procedures -- then they are generating garbage, that other people might accidentally rely upon.
Showing people don't actually follow the scientific method, is pretty much as good as showing the scientific method doesn't work.
Not the method is not valid, BUT: because people can't (or don't) actually follow the method, even when they say they do: AND, there's no method available of keeping people honest.
At least with the proper doing-up of seatbelts, there are police officers, to monitor, and enforce it in some cases (people still often don't wear their seatbelts) -- with scientific method implementation - not only is there no enforcement, but there are practical barriers that make enforcement, or reliable fraud detection, essentially impossible.