Comparing observational abundance features with nucleosynthesis predictions of stellar evolution or explosion simulations, we can scrutinize two aspects: (a) the conditions in the astrophysical production site and (b) the quality of the nuclear physics input utilized. We test the abundance features of r -process nucleosynthesis calculations for the dynamical ejecta of neutron star merger simulations based on three different nuclear mass models: The Finite Range Droplet Model, the (quenched version of the) Extended Thomas Fermi Model with Strutinsky Integral, and the Hartree–Fock–Bogoliubov mass model. We make use of corresponding fission barrier heights and compare the impact of four different fission fragment distribution models on the final r - process abundance distribution. In particular, we explore the abundance distribution in the second r -process peak and the rare-earth sub-peak as a function of mass models and fission fragment distributions, as well as the origin of a shift in the third r -process peak position. The latter has been noticed in a number of merger nucleosynthesis predictions. We show that the shift occurs during the r -process freeze-out when neutron captures and β -decays compete and an ( n , γ )–( γ , n ) equilibrium is no longer maintained. During this phase neutrons originate mainly from fission of material above A = 240. We also investigate the role of β -decay half-lives from recent theoretical advances, which lead either to a smaller amount of fissioning nuclei during freeze- out or a faster (and thus earlier) release of fission neutrons, which can (partially) prevent this shift and has an impact on the second and rare-earth peak as well.