It is shown that the space fluctuations of concentration
of conducting inclusions might be responsible for the well-known
disagreement between theory and experiment at determining microwave losses
in metal-dielectric mixture: the theories (percolation theory, effective
medium theory, etc.) predict much lower losses than those measured in
experiment. It is demonstrated that if the effective skin depth in the
regions occupied by the fluctuation is comparable to the mean diameter of
these regions we can expect additional losses.