The relative abundances of compound classes (e.g. CHO1, CHO2 and CHNO1) in fossil fuels have been widely used in previous researches based on the data from mass spectrometry (MS). However, such observations are susceptible to the ionization bias of MS due to the variation of ionization efficiency, which may affect the downstream quantitative analysis. Many efforts have been devoted to investigating the ionization efficiencies of a variety of model compounds, but not complex and diverse mixtures. Therefore, it is necessary to investigate how quantitatively ionization bias affects the MS analysis of complex mixtures. In this study, ionization bias was preliminarily investigated by comparing the elemental compositions obtained from MS-analysis via different ionization methods with the data acquired by a standard elemental analyzer. A coal-derived liquid and mixtures of five model compounds were ionized by atmospheric-pressure chemical ionization (APCI), electrospray ionization or direct analysis in real-time, followed by analyzing using a high-resolution Orbitrap MS to obtain elemental compositions. MS-based results via APCI ion source showed the closest elemental compositions to those obtained from the elemental analyzer. For the MS-based analysis, the abundances of S-and O-containing compounds were significantly lower and that of N-containing compounds was clearly higher compared to ultimate analysis. The abundances of compound classes were adjusted to provide a closer approximation of the actual contents, which might better serve for the downstream quantitative analysis. Although absolute quantification still needs additional work, the adjusted MS-based analysis provides not only the molecular level data, but also a better quantification ability compared to previous analytical approaches.