If you do an image search for "dunning kruger", you'll find an ocean of images like this one:

The effect as it is typically described: the most confident people are those who know just a little. 

Actual data gathered in these studies looks more like this:

The linear ordering is preserved, but compressed to 'slightly good'.

So when people casually discuss Dunning-Kruger, I think they would be better off just describing hubris. There's also a thing where people who are quite good underestimate themselves a little, but that could be explained in measurement error – both in how the experimenters determine skill, and how the people themselves determine their skill. I wouldn't worry about that bit too much, unless you want to point out a false humility.