We have selected a complete sample of flat-spectrum radio quasar from the WMAP 7-yr catalog, all with measured redshift within the SDSS area and have compared black hole mass estimates based on fitting a standard accretion disk model to the `blue bump' with with those obtained from the commonly used single epoch virial method. The sample comprises 79 objects flux limited at 1 Jy at 23 GHz, 52 of which (66\%) have a clearly detected `blue bump'. Thirty-three of the 52 FSRQ have, in the literature, black hole mass estimates obtained with the virial method. The mass estimates obtained from the two methods are well correlated. If the calibration factor of the virial relation is set $f=4.8$ , well within the range of recent estimates, the mean logarithmic ratio of the two mass estimates is equal to zero with a dispersion close to the estimated uncertainty of the virial method. The fact that the two totally independent methods agree so closely in spite of all the potentially large uncertainties associated to each of them lends strong support to both of them. The distribution of black-hole masses for the 52 FSRQs in our sample with a well detected blue bump has a median value of 10$^{9.0}$~M$_\odot$. It declines at the low mass end, consistent with other indications that radio loud AGNs are generally associated to the most massive black holes, although the decline may be, at least partly, due to the source selection. The distribution drops above $\log(M_\bullet/M_\odot) = 9.6$, implying that ultra-massive black holes associated to FSRQs must be rare.