The semiparametric accelerated hazard relational models proposed by Cadena and Denuit (Insurance Math. Econ. 68:1–16, 2016) are built from a reference life table by distorting the age scale. Individuals are made younger or older before performing the computations with the reference life table. In this paper, the mortality of individuals with higher educational attainment is modeled by means of an accelerated hazard relation specification. It is then projected by reference to the general population, using two different approaches. Numerical illustrations are performed on US mortality statistics.