At the end of the lanthanide series, 4f → 5d and other interconfigurational transitions, in which one electron is excited from a tight 4f orbital to a much more diffuse one, occur with a break of many f-f pairs, which make the electron correlation effects dominant. For instance, the large energy gap of 25 000 cm(-1) (∼29 500 cm(-1) without spin-orbit coupling) above the 4f(14) ground state of the SrCl2:Yb(2+) material is mostly due to electron correlation. In effect, a minimal multiconfigurational restricted active space (RASSCF) calculation that includes only the 4f(14) ground and 4f(13)5d and 4f(13)6s open-shell excited configurations gives a very small gap (5400 cm(-1)), whereas the correlation corrections to the 4f(14) → 4f(13)5d(eg) transition energies at the second order perturbation theory (RASPT2) level are very large: 35 599 ± 439 cm(-1), in average, for all excited states. These corrections are too large to be accurate at second order perturbation level. When a second f-shell is also included in the active space and single and double excitations to the 5d, 6s, and 5f shells are treated variationally, the (extended) RASSCF energy gap above the ground state and the electronic transitions increase by 22 038 ± 120 cm(-1) and the RASPT2 correlation energy corrections become small (-721 ± 571 cm(-1)), as it is desirable for a second order perturbation. A comparative analysis of both RASPT2 results reveals that the lack of the second f-shell accounts for 12 700 cm(-1) of the 14 223 ± 80 cm(-1) overestimation of interconfigurational transitions energies by the minimal RASPT2 calculation, which indicates an inaccurate calculation of the differential radial correlation between the 4f(14) and 4f(13)5d configurations by second order perturbation theory. In order to establish practical and accurate procedures for the calculation of 4f → 5d and other interconfigurational transitions at the end of the lanthanide series, the above and other RASSCF/RASPT2 calculations on the ionization potential of Yb(2+) in gas phase and in SrCl2 have been benchmarked in this paper against coupled cluster (coupled cluster singles and doubles and triples ) calculations, and RASSCF/RASPT2 calculations on the absorption spectrum of SrCl2:Yb(2+) have been compared with experiment. The results support that variational calculation of SD 4f → 5f excitations prior to RASPT2 calculations can be a realistic, accurate, and feasible choice to model radial correlation effects at the end of the lanthanide series.