Accounting for differences in element size and homogeneity when comparing finite element models : armadillos as a case study