Complete experimental studies of the production and transport of core and Rydberg nl states have been performed for Ar17+ at high velocity (v=23 a.u.) on different carbon targets; a range of target thicknesses from single collision condition to equilibrium has been investigated. Considering the collision system chosen, two models, based on a collisional approach -a rate-equation model and a classical Monte Carlo simulation- can be applied and compared. Our calculations show that they both represent well the experimental Rydberg l state populations but fail to describe the core states (n=2,3,4). The observed l-mixing for these states is much larger and faster than predicted. This discrepancy could possibly be explained by the Stark mixing induced by the projectile polarization of the medium, and neglected, thus far, in these two models.