I don't think optical constant can be taken as average of all constituent materials. Its like length of different parts of a sample can be average and that average can give the perception of overall length of sample But on the other hand, Density of different constituents can not be averaged to find the overall density of sample (unless we find volume of each constituent and multiply it with its density to find the mass of each constituent, add all the masses and divide by whole volume of the sample that will give us density of the sample).. our optical constant is a quantity and behave like density in the given analogy. By seeing the nature of Maxwell Garnett approx. i still doubt it will work or not ...
Unfortunately there is no winer in this competition. Each formula works for composite that has the same percolation threshold a this formula predicts. Real value of percolation threshold depends on micro correlations and can very from zero to unity.