I used different kind of fitting: once with local and once with parametric.
With local I got 58 up-regulated DE genes and 11 down-regulated, with parametric 13 up and 1 down.
There are two common up regulated genes and 1 down regulated.
I read that when there is a decreasing trend, then one should use a parametric fit but it gives me a small amount of DE genes somehow.
That are my plots for the dispersion.
local:

 

parametric:

Both fits look good to me.

Here is my general advice on why and how #DEG change when settings are changed:

A: Deseq2 results discrepancies because of versions of R