Stochastic weather generators combined with hydrological models have been proposed for continuous synthetic simulation to estimate return periods of extreme floods. Yet, this approach relies upon the length and spatial distribution of the precipitation input data series, which often are scarce, especially in arid and semiarid regions. In this work, we present a new approach for the estimation of extreme floods based on the continuous synthetic simulation method supported with inputs of (a) a regional study of extreme precipitation to improve the calibration of the weather generator (GWEX), and (b) non-systematic flood information (i.e., historical information and/or palaeoflood records) for the validation of the generated discharges with a fully distributed hydrological model (TETIS). The results showed that this complementary information of extremes allowed for a more accurate implementation of both the weather generator and the hydrological model. This, in turn, improved the flood quantile estimates, especially for those associated with return periods higher than 50 years but also for higher quantiles (up to approximately 500 years). Therefore, it has been proved that continuous synthetic simulation studies focused on the estimation of extreme floods should incorporate a generalized representation of regional extreme rainfall and/or non-systematic flood data, particularly in regions with scarce hydrometeorological records.