Description of ionization in the molecular approach to atomic collisions. II

We complement a previous article [Harel et al., Phys. Rev. A 55, 287 (1997)] that studied the characteristics of the description of ionization by the molecular approach to atomic collisions, by comparing the wave functions with accurate counterparts. We show how the failure of the basis to describe the phase of the ionizing wave function results in a trapping of the corresponding population in some molecular channels. The time evolution of the molecular wave function then departs from the exact one and the ionization and capture mechanisms appear as interlocked. We thus elucidate the question of the ``natural'' boundary of the molecular approach and draw further consequences as to the choice of pseudostates and the use of translation factors.