With this paper we provide a solution to a disagreement between astronomical- and laboratory-based CO-ice spectroscopic data. In observations towards icy sources, the CO-ice stretching band comprises a prominent and broad ‘red component’ around 2136.5 cm−1. This feature is generally attributed to solid CO mixed in a hydrogen-bonded environment like H2O, but, as far as we are aware, laboratory spectra have not been able to fully reproduce this feature. Water-containing CO ice cannot reproduce the observed band position and bandwidth without simultaneously producing a shoulder at 2152 cm−1 (4.647 μm). This band, believed to originate from the interaction of dangling-OH bonds with CO, is not observed in astronomical spectra. Fraser et al. suggested that the 2152 cm−1 feature is suppressed in astronomical ices by blocking of dangling-OH bonds by other species such as CO2 and CH3OH. In the present paper, we test this hypothesis by a systematic spectroscopic study of different H2O:CO:CO2 mixtures. It is shown that even though the 2152 cm−1 band is suppressed in low-temperature spectra, the width and peak position of the red component cannot be reproduced at the same time. An ice mixture containing only CO and CH3OH, however, does reproduce the spectra at low temperatures, both in terms of peak position and width of the red component, and the 2152 cm−1 band does not appear. This indicates that CO may reside in water-poor (rather than water-rich) ice in space. The astrophysical implications are discussed.