The realization of X-ray telescopes with imaging capabilities in the hard (> 10 keV) X-ray band requires the adoption of optics with shallow (< 0.25 deg) grazing angles to enhance the reflectivity of reflective coatings. On the other hand, to obtain large collecting area, large mirror diameters (< 350 mm) are necessary. This implies that mirrors with focal lengths ≥10 m shall be produced and tested. Full-illumination tests of such mirrors are usually performed with onground X-ray facilities, aimed at measuring their effective area and the angular resolution; however, they in general suffer from effects of the finite distance of the X-ray source, e.g. a loss of effective area for double reflection. These effects increase with the focal length of the mirror under test; hence a "partial" full-illumination measurement might not be fully representative of the in-flight performances. Indeed, a pencil beam test can be adopted to overcome this shortcoming, because a sector at a time is exposed to the X-ray flux, and the compensation of the beam divergence is achieved by tilting the optic. In this work we present the result of a hard X-ray test campaign performed at the BL20B2 beamline of the SPring-8 synchrotron radiation facility, aimed at characterizing the Point Spread Function (PSF) of a multilayer-coated Wolter-I mirror shell manufactured by Nickel electroforming. The mirror shell is a demonstrator for the NHXM hard X-ray imaging telescope (0.3 - 80 keV), with a predicted HEW (Half Energy Width) close to 20 arcsec. We show some reconstructed PSFs at monochromatic X-ray energies of 15 to 63 keV, and compare them with the PSFs computed from post-campaign metrology data, self-consistently treating profile and roughness data by means of a method based on the Fresnel diffraction theory. The modeling matches the measured PSFs accurately.