And ideally, the speed of light is \$ \frac{c}{\sqrt{\epsilon_r}} \$
In coaxial cable, where all the EM field is confined within the same dielectric, this is true.
In microstrip, most of the EM field is found in the dielectric, but a significant fraction is found in the air (or other upper dielectric) above the dielectric surface. Therefore the propagation velocity can't be calculated with this simple formula. The actual velocity will be slightly higher (very nearly an average of the velocity in air and the velocity in the dielectric, weighted by the proportion of the signal energy found in each).
You can use the result of your simulation to find the actual propagation velocity in your microstrip structure, and then re-tune the stub length to get the null at the frequency you want.
Qucs should include a microstrip open stub element, which will also model fringing effects at the open end of the stub line. These will also slightly change the resonant frequency of the stub. (Meaning, your current simulation won't quite be accurate due to not including these effects)