It is often envisioned that cations might coordinate at specific sites of nucleic acids and play an important structural role, for instance in the transition between B-DNA and Z-DNA. However, nucleic acid models explicitly devoid of specific sites may also exhibit features previously considered as evidence for specific binding. Such is the case of the "composite cylinder" (or CC) model which spreads out localized features of DNA structure and charge by cylindrical averaging, while sustaining the main difference between the B and Z structures, namely the better immersion of the B-DNA phosphodiester charges in the solution. Here, we analyze the non-electrostatic component of the free-energy difference between B-DNA and Z-DNA. We also compute the composition of the counterion sheath in a wide range of mixed-salt solutions and of temperatures: in contrast with the large difference of composition between the B-DNA and Z-DNA forms, the temperature dependence of sheath composition, previously unknown, is very weak. In order to validate the model, the mixed-salt predictions should be compared to experiment. We design a procedure for future measurements of the sheath composition based on Anomalous Small-Angle X-ray Scattering and complemented by P-31 NMR. With due consideration for the kinetics of the B-Z transition and for the capacity of generating at will the B or Z form in a single sample, the 5'-d[T-(m(5)C-G)(12)-T] 26-mer emerges as a most suitable oligonucleotide for this study. Finally, the application of the finite element method to the resolution of the Poisson-Boltzmann equation is described in detail. (C) 2016 Wiley Periodicals, Inc.