I think whether you can say it's empirically incorrect depends on if the researchers measured the right thing. Measuring: “Do you think your job is useful?” was over-simplifying, and could miss the phenomenon that Graeber was actually talking about.