DFT is an approximation. It's good for some molecules and it's bad for others. There are many methods under the DFT umbrella, so it's more complicated. Some method are good for some molecules, other methods are good for others. Some molecules are easy and can be approximated by many methods, some molecules are hard and no method is good for them.

It's not my area, but I'm sure B80 is one of the tricky ones. In general anything with Boron is hard. This in particular probably have some electrons that are not inside a "bond" between two atoms, but are distributed in the whole molecule. Something like benzene, that has a few electrons in a circular ring of 6 atoms, but in this case it's 3D and with 80 atoms. You need some special cases for the ring in benzene and similar molecules.

The main problem is that solving the molecules exactly needs exponential time in a classical computer. If H is the number of Hydrogen and X is the number of very light atoms, it's like (expt(2*(H+5X)))^3. Heavy atoms enter with a bigger multiplier. And that bound already has a lot approximations and simplifications. So for not trivial molecules and for big molecules that are important in biology with X~=100 or 1000 you must do some approximations.

DFT is one of them. Most of the time it works, specially if you choose the correct method inside the DFT label. I'm not surprised that there are exceptions. If confirmed, probably someone will create a new tweak inside of one of the method to fix the discrepancy.