There are major discrepancies between recent intermediate coupling frame transformation (ICFT) and Dirac atomic R-matrix code (DARC) calculations (Fernández-Menchero et al 2014 Astron. Astrophys. 566 A104; Aggarwal et al 2016 Mon. Not. R. Astron. Soc. 461 3997) regarding electron-impact excitation rates for transitions in several Be-like ions, as well as claims that the DARC calculations are much more accurate and the ICFT results might even be wrong. To identify possible reasons for these discrepancies and to estimate the accuracy of the various results, we carried out independent B-spline R-matrix calculations for electron-impact excitation of the Be-like ion N3+. Our close-coupling (CC) expansions contain the same target states (238 levels overall) as the previous ICFT and DARC calculations, but the representation of the target wave functions is completely different. We find close agreement among all calculations for the strong transitions between low-lying states, whereas there remain serious discrepancies for the weak transitions as well as for transitions to highly excited states. The differences in the final results for the collision strengths are mainly due to differences in the structure description, specifically the inclusion of correlation effects, rather than the treatment of relativistic effects or problems with the validity of the three methods to describe the collision. Hence there is no indication that one approach is superior to another, until the convergence of both the target configuration and the CC expansions have been fully established.