Earth-based radar observations of the rotational dynamics of Mercury combined with the determination of its gravity field by MESSENGER (Smith et al. 2012) give clues on the internal structure of Mercury, in particular its polar moment of inertia C, deduced from the obliquity (2.04 ± 0.08) arcmin.The dynamics of the obliquity of Mercury is a very-long term motion (a few hundreds of kyrs), based on the regressional motion of Mercury's orbital ascending node. This paper, following the study of Noyelles & D'Hoedt (2012), aims at first giving initial conditions at any time and for any values of the internal structure parameters for numerical simulations, and at using them to estimate the influence of usually neglected parameters on the obliquity, like J 3 , the Love number k 2 and the secular variations of the orbital elements. We use, for that, averaged representations of the orbital and rotational motions of Mercury, suitable for long-term studies.We find that J 3 should alter the obliquity by 250 milli-arcsec, the tides by 30 milli-arcsec, and the secular variations of the orbital elements by 10 milliarcsec over 20 years. The resulting value of C could be at the most changed from 0.346mR 2 to 0.345mR 2 .