A reasonable estimate of the resolution of a point cloud obtained by Structure-from-Motion (SfM) photogrammetry is mandatory in order to perform a correct analysis of the detected object. The knowledge of the expected resolution and of its limits allows one to establish, right from the measurement design stage, realistic hypothesis on the achievable degree of detail. This paper shows a case study designed for applications in the architectural field, where the distances involved generally range from few meters to a few tens of meters. Therefore, an artificial 3D target composed of both flat and three-dimensional elements was devised and measured in order to evaluate the degree of perturbation in a 3D case with respect to ideal 2D visibility conditions. The results confirmed and extended what already evidenced in previous publications, showing that resolution gets worst of about 20-25% when compared with the 2D case.