Context. Massive stars play a fundamental role in the evolution of the Universe, yet several physical processes governing their evolution remain poorly constrained. Notably, the main-sequence width is known to be sensitive to the convective boundary mixing efficiency. Moreover, discrepancies between single-star evolution predictions and observations highlight the need to account for binary interactions to explain some populations of massive stars. Aims. Our goal is to constrain single-star models using recent observations of massive Galactic stars from the IACOB database. We use the latest proposed empirical location of the terminal age main sequence (TAMS) to calibrate the convective boundary mixing efficiency and use this calibration to test single-star evolution by comparing various model predictions to the observed populations of the IACOB sample. Methods. We compute detailed single-star models with GENEC. We calibrate the convective boundary mixing efficiency, ensuring that the models reproduce the proposed empirical TAMS location. We compute several grids with various overshoot calibrations, angular momentum (AM) transport treatments, initial masses (Mini = 12 - 40 M⊙) and velocities (υini/υcrit = 0.0 - 0.6). Finally, we generate synthetic populations from the tracks with SYCLIST and perform a direct comparison with the observed population. Results. The calibrated models at slow rotation (υini/υcrit = 0.1) reproduce the empirical TAMS location. We find that a mass-dependent overshoot efficiency is required to fit the observational constraints. Additionally, the overall rotational properties of the observed populations are well reproduced with single-star models, independently of the AM transport assumptions. In particular, models accounting only for hydrodynamical instabilities are successful at reproducing the rotational properties, unlike previous GENEC grids, which we attribute to the choice of wind prescription. Although the empirical TAMS of slow rotators (υ sin i < 100 km/s) is well reproduced, we find that models are unsuccessful at explaining the velocity dependence of the TAMS location observed in the IACOB sample. Finally, we find that single-star models fail at explaining the population of blue supergiants to the right of the TAMS location, as they cross the Hertzsprung gap too fast. The synthetic populations strengthen this result, predicting ~2% of stars in this region instead of the observed ~15%.