The Large Magellanic Cloud (LMC) star cluster system offers the unique opportunity to independently check the accuracy of age (and the corresponding mass) determinations based on a number of complementary techniques. Using our sophisticated tool for star cluster analysis based on broad-band spectral energy distributions (SEDs), 'AnalySED', we re-analyse the Hunter et al. LMC cluster photometry. Our main aim is to set the tightest limits yet on the accuracy of absolute age determinations based on broad-band SEDs, and therefore on the usefulness of such an approach. Our broad-band SED fits yield reliable ages, with statistical absolute uncertainties within Delta log(Age/yr) similar or equal to 0.4 overall. The systematic differences we find with respect to previous age determinations are caused by conversions of the observational photometry to a different filter system, thus leading to systematically inaccurate results. The LMC's cluster formation rate (CFR) has been roughly constant outside of the well-known age gap between similar to 3 and 13 Gyr, when the CFR was a factor of similar to 5 lower. Using a simple approach to derive the characteristic cluster disruption time-scale, we find that log(t(4)(dis)/yr) = 9.9 +/- 0.1, where t(dis)=t(4)(dis)(M-cl/10(4) M-circle dot)(0.62). This long characteristic disruption time-scale implies that we are observing the initial cluster mass function (CMF). We conclude that while the older cluster (sub)samples show CMF slopes that are fully consistent with the alpha similar or equal to-2 slopes generally observed in young star cluster systems, the youngest mass and luminosity-limited LMC cluster subsets show shallower slopes (at least below masses of a few x10(3) M-circle dot), which is contrary to dynamical expectations. This may imply that the initial CMF slope of the LMC cluster system as a whole is not well represented by a power law, although we cannot disentangle the unbound from the bound clusters at the youngest ages.