In this article, we analytically present the influence of the systematic error due to the theoretical assumptions on the blood oxygenation determination using a 'Phase-only' dual-wavelength phase-modulation spectroscopy (PMS). To approach an accurate quantification of hemoglobin saturation, an empirical method to experimentally acquire the initial phase-shift value (theta(0)(lambda)) in oxygenation quantification from PMS data is investigated on blood-perfused phantom. Our results show that the quantification error of blood oxygenation determination can be substantially compensated by using the acquired theta(0)(lambda) method and is confirmed to be less than 10% by comparison with the measured results of the Pt electrode.