In the past decades various algorithms have been developed for the retrieval of water constituents from the measurement of ocean color radiometry, and one of the approaches is the spectral optimization scheme. This approach defines a target function (or error function) between the input remote sensing reflectance and the output remote sensing reflectance, with the latter modeled with a few variables that represent the optically active properties (such as the absorption coefficient of phytoplankton and the backscattering coefficient of particles). The values of the variables are considered the water properties that form the input remote sensing reflectance when the error function reaches a minimum (optimization is achieved); in other words, because the remote sensing reflectance of many spectral bands are used in the process and each band forms an equation, the series of equations are solved simultaneously and numerically. The applications of this approach implicitly assume that the error function is monotonic of the various variables. Here, the error function is proven theoretically that it has no local minimum as long as the number of variables is kept small. In addition, examples of the shape of the error surface are also presented with numerically simulated data, in a way to justify the possibility of finding a solution of the various variables. More importantly, because the spectral properties could be modeled differently, the impacts of such differences on the error surface as well as on the retrievals are also presented. The results suggest that the various spectral models are not equal in ocean-color remote sensing although all use an optimization approach.