A natural interpretation of the correlation between nearby Active Galactic Nuclei (AGN) and the highest-energy cosmic rays observed recently by the Pierre Auger Collaboration is that the sources of the cosmic rays are either AGN or other objects with
a similar spatial distribution (the ``AGN hypothesis). We question this interpretation. We calculate the expected distribution of the arrival directions of cosmic rays under the AGN hypothesis and argue that it is not supported by the data, one of manifestations of the discrepancy being the deficit of events from the direction of the Virgo supercluster. We briefly discuss possible alternative explanations including the origin of a significant part of the observed events from Cen A.
We analyse a sample of 33 extensive air showers (EAS) with estimated primary energies above 2cdot 10^{19} eV and high-quality muon data recorded by the Yakutsk EAS array. We compare, event-by-event, the observed muon density to that expected from COR
SIKA simulations for primary protons and iron, using SIBYLL and EPOS hadronic interaction models. The study suggests the presence of two distinct hadronic components, ``light and ``heavy. Simulations with EPOS are in a good agreement with the expected composition in which the light component corresponds to protons and the heavy component to iron-like nuclei. With SYBILL, simulated muon densities for iron primaries are a factor of sim 1.5 less than those observed for the heavy component, for the same electromagnetic signal. Assuming two-component proton-iron composition and the EPOS model, the fraction of protons with energies E>10^{19} eV is 0.52^{+0.19}_{-0.20} at 95% confidence level.
