The 146Sm/144Sm ratio in the early solar system has been constrained by Nd/Sm isotope ratios in meteoritic material. Predictions of 146Sm and 144Sm production in the γ process in massive stars are at odds with these constraints, and this is partly due to deficiencies in the prediction of the reaction rates involved. The production ratio depends almost exclusively on the (γ,n)/(γ,α) branching at 148Gd. A measurement of 144Sm(α,γ)148Gd at low energy had discovered considerable discrepancies between cross-section predictions and the data. Although this reaction cross section mainly depends on the optical α+nucleus potential, no global optical potential has yet been found that can consistently describe the results of this and similar α-induced reactions at the low energies encountered in astrophysical environments. The untypically large deviation in 144Sm(α,γ) and the unusual energy dependence can be explained, however, by low-energy Coulomb excitation, which is competing with compound nucleus formation at very low energies. Considering this additional reaction channel, the cross sections can be described with the usual optical potential variations, compatible with findings for (n, α) reactions in this mass range. Low-energy (α, γ) and (α, n) data on other nuclei can also be consistently explained in this way. Since Coulomb excitation does not affect α emission, the 148Gd(γ,α) rate is much higher than previously assumed. This leads to very small 146Sm/144Sm stellar production ratios, in even more pronounced conflict with the meteorite data.