Amplitude modulation atomic force microscopy is one of the most broadly used techniques for the nanoscale characterization of a large variety of surfaces because it can routinely provide topography images with nanometer and subnanometer resolution in air, i.e. under ambient conditions, using available commercial instruments. The topographic map results from the convolution of the different interactions (van der Waals, capillary, adhesion, etc.) sensed by the probe and the presence of nanometer-thick water films on both the surface and the tip of the probe, as is usually the case under ambient conditions, can lead to apparent heights markedly different from the real heights due to formation and rupture of water menisci, particularly when the surfaces exhibit regions with different affinity to water (hydrophilic and hydrophobic). In order to systematically explore such a well-known but usually ignored phenomenon, we have performed a combined experimental and theoretical study using (hydrophobic) self-assembled monolayers of stearic acid grown on (hydrophilic) freshly cleaved mica surfaces and a simplified point mass on a spring model to simulate the tip dynamics. We show that, depending on the operation parameters (free oscillation amplitude and setpoint), the apparent heights can vary in magnitude and sign (contrast inversion) and, most important, that the true height cannot be measured in the presence of water layers when surface affinity to water is not homogeneous even if menisci are not formed. We suggest to revise, within the perspective of the present investigation, those published works where the determination of heights is critical.