You've compared latent heat of vapourisation with the liquid freezing energy. There's a slight difference.
Having run several freezing models over the years the solver is usually fairly good. What tends to cause problems are mesh resolution, mesh (and cell) quality and material/boundary settings. Time step can also play a part if the frozen region causes flow patterns to change and then the frozen parts to melt as a result of this.
You are navigating away from the AIS Discovery experience