We have analyzed the efficiency in source detection and flux density estimation of blind and non-blind detection techniques exploiting the MHW2 filter applied to the Wilkinson Microwave Anisotropy Probe (WMAP) 5-year maps. A comparison with the AT20G
Bright Source Sample (Massardi et al. 2008), with a completeness limit of 0.5 Jy and accurate flux measurements at 20 GHz, close to the lowest frequency of WMAP maps, has allowed us to assess the completeness and the reliability of the samples detected with the two approaches, as well as the accuracy of flux and error estimates, and their variations across the sky. The uncertainties on flux estimates given by our procedure turned out to be about a factor of 2 lower than the rms differences with AT20G measurements, consistent with the smoothing of the fluctuation field yielded by map filtering. Flux estimates were found to be essentially unbiased except that, close to the detection limit, a substantial fraction of fluxes are found to be inflated by the contribution of underlying positive fluctuations. This is consistent with expectations for the Eddington bias associated to the true errors on flux density estimates. The blind and non-blind approaches are found to be complementary: each of them allows the detection of sources missed by the other. Combining results of the two methods on the WMAP 5-year maps we have expanded the non-blindly generated New Extragalactic WMAP Point Source (NEWPS) catalogue (Lopez-Caniego et al. 2007) that was based on WMAP 3-year maps. After having removed the probably spurious objects not identified with known radio sources, the new version of the NEWPS catalogue, NEWPS_5yr comprises 484 sources detected with a signal-to-noise ratio SNR>5.
