It is found that a significant spread in the optical constants of metals reported by different authors is caused by differences in the sample preparation methods, measurement conditions, and methods of calculation of sought parameters, as well as by the oxidation effect. It is shown that the optical constants of metals in films 80–120 nm thick on silicon substrates with scattering below 10−4 are determined with minimal errors. The reflectance of these mirrors calculated from the optical constants found by the most accurate ellipsometric method coincides with the experimental value within the measurement accuracy. Low values of k(λ) obtained for thin layers in some works using the methods based on the measurement of the coherent transmittance and regular reflectance are explained by disregarded scattering and luminescence. The spectra of the imaginary part of the complex refractive index of copper, nickel, and copper oxide determined by us by the proposed methods for thin nanostructured layers taking into account the scattering and luminescence coincide with the most correct data for thick films in the spectral range of 325–633 nm. For thin palladium and palladium oxide layers, the variations in k(λ) are caused by the oxidation of metal granules and disregarded luminescence for thick oxide layers in the long-wavelength spectral region. The maximal difference in the imaginary part of the complex refractive index of copper and nickel for thin nanostructured layers are observed in the region of plasmon resonances, whose positions and amplitudes depend on the degree of asphericity, the shape, and the degree of order of particles and their aggregates, which shift the plasmon resonances of films to longer wavelengths with respect to spherical particles.