Chances are that the balance between the reference spectrum and sample spectrum in terms of degree of purge is not correct, if you use atmospheric drier since it gets rid of water and carbon dioxide. The 2363 cm-1 is likely due to the atmospheric CO2 in the spectrometer. Even if you did not use the drier, if the path length of the reference and sample are different, then you will see the difference in CO2 intensity that then appears in the final spectrum. See if the peak position of this agrees with the very strong band by simply recording a single beam spectrum of air (not a transmittance).