The new generation of spaceborne hyperspectral sensors offers the potential to provide new information on water quality, especially regarding phytoplankton groups. While phytoplankton species composition algorithms are easily demonstrated in model studies, the application of algorithms to spaceborne data may be much more challenging because of processing artefacts in the spectral vicinity of atmospheric absorption features. If not treated carefully, the unresolved spectral variability of radiance and irradiance can create spectral artefacts ("wiggles") in water reflectance spectra with high second derivative and thus contaminates pigment detection algorithms. In this study one of the processes generating such spectral wiggles is explained for an in situ radiometer system with a wavelength offset between the radiance and irradiance measurements. The two measurements used to calculate reflectance are differently affected by narrow, unresolved atmospheric absorption bands. Removal or reduction of such wiggles could be achieved by physically-based weighting in the spectral interpolation of irradiance rather than the typical spectral averaging generally used to hide such problems when using linear interpolation. Although demonstrated here for in situ radiometry the need for careful treatment of unresolved spectral variability in spaceborne data is raised, particularly if such data is used for second derivative algorithms which are very sensitive to short wavelength scale variability.