Hi Salvador,
They really can produce very different results. My simulations are very conservative in terms of the choices of covariates and this still showed up. I was a bit surprised to hear that there was a group of people that considered the choice inconsequential. Model selection matters in this case.