Idealized models of cylinders for the vasculature are used in several quantitative MRI techniques such as for perfusion, CBV, vessel size and vascular MR fingerprinting. While limitations of these models are recognized, a direct comparison of the predicted MR signal between different cylinder models and those using a real vasculature as substrate has not been done to our knowledge. Here we compare the sensitivity of the MR signal for the GESFIDE sequence for 4 sets of , models of cylinders and 6 realistic angiograms from mouse somatosensory cortex. In general, simulation results are all different between the different angiograms and the different models of cylinders. This suggests that much care should be used in interpreting literature results based on models of cylinders, or as well with models on angiograms, to account for the possibility of biases in the absolute results. Correlations and differences in absolute values, for some parameters, may perhaps be less subject to bias.