Tissue oxygen saturation (StO2) is a valuable clinical parameter e.g. for intensive care applications or monitoring during surgery. Studies showed that near-infrared spectroscopy (NIRS) based tissue oximeters of different brands give systematically different readings of StO2. Usually these readings are linearly correlated and therefore StO2 readings from one instrument can easily be converted to those of another instrument. However, it is interesting to understand why there is this difference. One reason may be that different brands employ different spectra of hemoglobin. The aim here was to investigate how these different absorption spectra of hemoglobin affect the StO2 readings. Therefore, we performed changes in StO2 in a phantom experiment with real human hemoglobin at three different concentrations (26.5, 45 and 70 μM): desaturation by yeast consuming the oxygen and re-saturation by bubbling oxygen gas. The partial pressure of O2 in the liquid changed from at least 10 kPa to ~0 kPa and ISS OxiplexTS, a frequency-domain NIRS instrument, was used to monitor changes of StO2. When we employed two different absorption spectra for hemoglobin, StO2 values were comparable in the normal physiological range. However, particularly at high and low StO2 values, a difference of >6% between these two spectra were noticed. Such a difference of >6% is substantial and relevant for medical applications. This may partly explain why different brands of NIRS instruments provide different StO2 readings. The hemoglobin spectra are therefore a factor to be considered for future developments and applications of NIRS oximeters.