Energies from the GW approximation and the Bethe-Salpeter equation (BSE) are benchmarked against the excitation energies of transition-metal (Cu, Zn, Ag, and Cd) single atoms and monoxide anions. We demonstrate that best estimates of GW quasiparticle energies at the complete basis set limit should be obtained via extrapolation or closure relations, while numerically converged GW-BSE eigenvalues can be obtained on a finite basis set. Calculations using real-space wave functions and pseudopotentials are shown to give best-estimate GW energies that agree (up to the extrapolation error) with calculations using all-electron Gaussian basis sets. We benchmark the effects of a vertex approximation (Γ) and the mean-field starting point in GW and the BSE, performing computations using a real-space, transition-space basis and scalar-relativistic pseudopotentials. While no variant of GW improves on perturbative GW at predicting ionization energies, GWΓ-BSE computations give excellent agreement with experimental absorption spectra as long as off-diagonal self-energy terms are included. We also present GW quasiparticle energies for the CuO, ZnO, AgO, and CdO anions, in comparison to available anion photoelectron spectra.