We study whether dry merger-driven size growth of massive elliptical galaxies depends on their initial structural concentration, and analyse the validity of the homology hypothesis for virial mass determination in massive ellipticals grown by dry mergers. High-resolution simulations of a few realistic merger trees, starting with compact progenitors of different structural concentrations (Sérsic indices n), show that galaxy growth has little dependence on the initial Sérsic index (larger n leads to slightly larger size growth), and depends more on other particulars of the merger history. We show that the deposition of accreted matter in the outer parts leads to a systematic and predictable breaking of the homology between remnants and progenitors, which we characterize through the evolution, during the course of the merger history, of virial coefficients K≡ G M / R_e σ _e^2 associated with the most commonly used dynamical and stellar mass parameters. The virial coefficient for the luminous mass, K⋆, is ∼50 per cent larger at the start of the merger evolution at z ≈ 2 than in z = 0 remnants. Ignoring virial evolution leads to biased virial mass estimates. We provide K corresponding to a variety of dynamical and stellar mass parameters, and provide recipes for the dynamical determination of galaxy masses. For massive, non-compact ellipticals, the popular expression M = 5 R_e σ _e^2 / G underestimates the dynamical mass within the luminous body by factors of up to 4; it instead provides an approximation to the total stellar mass with smaller uncertainty than current stellar-population models.