Near-field scanning microwave microscopy (NSMM) has to face several issues for the establishment of traceable and quantitative data. In particular, at the nanoscale, the wavelength of operation in the microwave regime appears disproportionate compared to the size of the nano-object under investigation. Incidentally, the microwave characterization results in poor electrical sensitivity as the volume of the wave/material interaction is limited to a fraction of the wavelength. In addition, the definition of nanoscale microwave impedance standards requires accurate knowledge of the material and dimensional properties at such scale. In this effort, a quantitative error analysis performed on micrometric metal oxide semiconductor (MOS) structures is proposed. In particular, atomic force microscopy (AFM) image together with the magnitude and phase-shift images of the complex microwave reflection coefficient using a Keysight TM 's LS5600 AFM interfaced directly with a vector network analyzer, without electrical matching strategy, are performed around 9.5GHz. From a detailed analysis of the raw data, completed with a FEM-based electromagnetic modeling, quantitative capacitances extraction and system limitations are exemplary shown. Keywords-near-field scanning microwave microscopy (NSMM), vector network analyzer (VNA), metal oxide semiconductor (MOS), vector calibration.