I'd note that the figure 2 you linked is the one based on logical reasoning and thus probably the most key figure and that it
does show less competent people rating themselves higher than more competent people - although not higher than the most competent.
Personally, I think the most significant part of the paper is study 3 where they found that the unskilled not only predicted their own skill wrong but were unable to re-calibrate when shown the results of the more skilled.
(The paper is
Available here for free if anyone wants to read it)