According to Callister pg 517, the rule of mixtures governs the strength of a large particle composite such as E/G or E/Q. Running the rule of mixtures for our US Composites #635 19000psi epoxy and some 10.5e6psi glass spheres, once can see that the mixture is not assured to achieve a strength more than several times that of the epoxy itself until well over 80% fill rate is achieved.
According to http://mathworld.wolfram.com/SpherePacking.html the packing density of glass spheres like zeospheres varies from between approximately 5% and 75%. At 5% spheres, the mix would be almost all epoxy and thus perform badly. The smaller spheres seem to fight this effect and pack in more uniform and dense arrangements.
Regardless of size, getting better than a 75% fill rate on glass spheres of a single size is mathematically impossible. This indicates that we need either multiple sizes of spheres or spheres and fibers as our macro reinforcements.
The variance in strength for a given ratio is huge. We can see from the graph that the most effective way of minimizing the variance and maximizing strength is to minimize the epoxy fraction as walter correctly predicted from his experiments. It seems to me that one of the other posters mentioned the second aggregate should be 1/5 the size of the first but I couldn't find the post with that reference.
What this says to me is that the correct aggregate mix is going to be something like 75% G800 zeospheres and 25% of something like G200 zeospheres. Zeospheres are nice if they aren't too expensive since they are quality controlled and thus likely to produce similar results each time.
This writeup neglects the effects of the exotic additives in my previous posts and just focuses on materials science that has been known for the last 50 years. In short, we first need to get the epoxy fraction down by using multiple aggregate sizes and then we can focus on the special property improving additives.