We calculate the capture (fusion) cross sections for nine reactions involving spherical nuclei: 16 O + 16 O, 28 Si, 92 Zr, 144 Sm, 208 Pb; 28 Si + 28 Si, 92 Zr, 208 Pb; 32 S + 208 Pb. For six of them precision data are available in the literature. Analysis of these precision data within the framework of the single-barrier penetration model based on the Woods-Saxon profile for the strong nucleus-nucleus interaction potential (SnnP) gave rise to the problem of the apparently large diffuseness of the SnnP [Newton et al., Phys. Rev. C 70, 024605 (2004)]. Our fluctuation-dissipation trajectory model is based on the double-folding approach with the density-dependent M3Y NN forces including the finite-range exchange part. For the nuclear matter density the Skyrme-Hartree-Fock approach including the tensor interaction is applied. The resulting nucleus-nucleus potential possesses rather small (normal) diffuseness. The strength of the radial friction K R is used as the free parameter of the model. It turns out that for four of the five reactions induced by 16 O (except 16 O + 208 Pb) the calculated cross sections cannot be brought into agreement with the data within the experimental errors. This suggests that the calculated nuclear density is incorrect for 16 O. For the reactions not involving 16 O and, surprisingly, for the 16 O + 208 Pb reaction the agreement with the data within 2-5% is achieved at K R = 1.2 × 10 −2 to 3.0 × 10 −2 MeV −1 zs which is in accord with the previous works.