Remote sensing of chlorophyll-a concentration, [Chl-a], has been difficult in coastal waters like the Chesapeake Bay owing largely to terrestrial substances (such as minerals and humus) that are optically significant but do not covary with phytoplankton. Here we revisit the semi-analytical pathway of deriving [Chl-a] based on the light absorption coefficient of phytoplankton by introducing the generalized stacked-constraints model (GSCM) to partition satellite-derived total light absorption coefficient of water (with pure-water contribution subtracted), anw(λ), into phytoplankton, aph(λ), and non-phytoplankton components, where anw(λ) is derived from satellite remote-sensing reflectance, Rrs(λ), using the Quasi-Analytical Algorithm. The GSCM-derived aph(λ) was compared with field matchups of [Chl-a]. We show that semi-analytical approaches can provide superior [Chl-a] product compared with reflectance-band-ratio algorithms when the accuracy of satellite-derived aph(λ) is sufficiently improved, in this case with the GSCM. However, the improvement is at the cost of significantly reduced data availability because the GSCM may provide no feasible solutions when input anw(λ) data are subject to large errors. This in turn highlights the needs for improved atmospheric correction and upstream models capable of preserving actual spectral shapes of Rrs(λ) and anw(λ), respectively.