Mid-infrared microspectroscopy is a non-invasive tool for identifying the molecular structure and chemical composition at the scale of the probe, i.e. at the scale of the beam. Consequently, investigating small objects or domains (commensurable to the wavelength) requires high-resolution measurements, even down to the diffraction limit. Herein, different protocols and machines allowing high-resolution measurements in transmission mode (aperture size (i.e., beam size) from 15 × 15 μm to 3 × 3 μm) are tested using the same sample. The model sample is a closed cavity containing a water-air assemblage buried in a quartz fragment (fluid inclusion). The spectral range covers the water stretching band (3000-3800 cm-1), whose variations are followed as a function of the distance to the cavity wall. The experiments compare the performance of one focal plane array (FPA) detector associated with a Globar source with respect to a single-element mercury cadmium telluride (MCT) detector associated with a supercontinuum laser (SCL) or a synchrotron radiation source (SRS). This work also outlines the importance of post-experimental data processing, including interference fringe removal and Mie scattering correction, to ensure that the observed spectral signatures are not related to optical aberrations. We show that the SCL and the SRS-based setups detect specific spectral features along the quartz boundary (solid surface), invisible to the FPA imaging microscope. Additionally, the broadband SCL thus has the potential to substitute at the laboratory scale the SRS for conducting diffraction-limited high-resolution measurements.