We analyze recent AMS-02 comic-ray measurements of lithium, beryllium, boron, carbon, nitrogen, and oxygen. The emphasis of the analysis is on systematic uncertainties related to propagation and nuclear cross sections. To investigate the uncertainties in the propagation scenario, we consider five different frameworks, differing with respect to the inclusion of a diffusion break at a few GV, the presence of reacceleration, and the presence of a break in the injection spectra of primaries. For each framework we fit the diffusion equations of cosmic rays and explore the parameter space with Monte Carlo methods. At the same time, the impact of the uncertainties on the nuclear production cross sections of secondaries is explicitly considered and included in the fit through the use of nuisance parameters. We find that all of the considered frameworks are able to provide a good fit. In particular, two competing scenarios—one including a break in diffusion but no reacceleration and the other with reacceleration but no break in diffusion—are both allowed. The inclusion of cross-section uncertainties is, however, crucial to this result. Thus, at the moment these uncertainties represent a fundamental systematic preventing a deeper understanding of the properties of cosmic-ray propagation. Nonetheless, we find that the slope of diffusion at intermediate rigidities is robustly constrained in the range δ≃0.45–0.5 in models without convection, or δ≃0.4–0.5 if convection is included in the fit. Furthermore, we find that the use of the AMS-02 beryllium data provides a lower limit on the vertical size of the Galactic propagation halo of zh≳3  kpc.