While referring to some of the papers on THz metamaterials, I found that effective medium theory has been applied to structures much larger than 1/4th of the wavelength.
I agree with Kumar that EMT can only be a crude guideline for most metamaterial applications. In my experience the wavelength (in the material) should be less than 1/10 to be trustworthy. Moreover many of the metamaterials are merely surfaces and one cannot estimate a bulk property.
Photonic crystals give a more trustworthy result, but they should include finite-size effects using, say, transfer matrix methods. Some time ago Halevi published rigorous results on EMT within the context of dielectric photonic crystals.
For surface limited metamaterials one may try RCWA; otherwise Comsol (FEM)calculations may yield reliable results.