Wang-Landau second code delivers too high and too low values for the density of states

[IreneFerri]

The second Wang-Landau program (which achieve flatness for any N value) delivers a wrong output. However, if we exclude very high and very low points from the graphic, we obtain a figure similar to the one produced by the exact calculation, performed by 'dos_mean.py' (see 'code' and 'images' folders), just in another scale. I have checked manually for N=3 if there was any issue related to the summation of degenerated energies but this part seems to work fine, so I guess the problem is in the Wang-Landau itself.

Comparing the results obtained by the 2 Wang-Landau codes and the exact calculation for N=14, we can observe that the first sampling matches the exact result (just need to substract the minimum to all values for identical results, and this readjustment is always necessary for Wang-Landau).