A meta-analysis can only be as good as the sum of its parts, and if its parts are very heterogeneous, which they often are, but not heterogeneous in the right way, then your analyses are somewhat limited.