We investigate the characteristics of bright flares for a sample of supergiant fast X-ray transients and their relation to the orbital phase. We have retrieved all Swift/BAT Transient Monitor light curves, and collected all detections in excess of $5sigma$ from both daily- and orbital-averaged light curves in the time range of 2005-Feb-12 to 2013-May-31. We also considered all on-board detections as recorded in the same time span and selected those within 4 arcmin of each source in our sample and in excess of $5sigma$. We present a catalogue of over a thousand BAT flares from 11 SFXTs, down to 15-150keV fluxes of $sim6times10^{-10}$ erg cm$^{-2}$ s$^{-1}$ (daily timescale) and $sim1.5times10^{-9}$ erg cm$^{-2}$ s$^{-1}$ (orbital timescale, averaging $sim800$s) and spanning 100 months. The great majority of these flares are unpublished. This population is characterized by short (a few hundred seconds) and relatively bright (in excess of 100mCrab, 15-50keV) events. In the hard X-ray, these flares last in general much less than a day. Clustering of hard X-ray flares can be used to indirectly measure the length of an outburst, even when the low-level emission is not detected. We construct the distributions of flares, of their significance (in terms of sigma) and their flux as a function of orbital phase, to infer the properties of these binary systems. In particular, we observe a trend of clustering of flares at some phases as $P_{rm orb}$ increases, as consistent with a progression from tight, circular or mildly eccentric orbits at short periods, to wider and more eccentric orbits at longer orbital periods. Finally, we estimate the expected number of flares for a given source for our limiting flux and provide the recipe for calculating them for the limiting flux of future hard X-ray observatories. (Abridged).