Spectral prediction models for halftone prints generally assume homogeneously thick and sharply edged ink dots, i.e., bilevel halftones. In real prints, the ink thickness often decreases at the boundaries of the ink dots, thereby forming continuous-level halftones. The present study aims at verifying to what extent the classical Clapper-Yule and Yule-Nielsen models are able to predict the reflectance of single-ink continuous-level halftone prints. First we model the reflectance of continuous-level halftones by developing variable thickness extensions of both the Clapper-Yule and the Yule-Nielsen spectral prediction models. We consider continuous halftones whose thickness profiles are obtained by Gaussian filtering of the bilevel halftone image. Then we predict the reflectance spectra defined by the continuous-level models by fitting the bilevel models' effective ink surface coverages. Since dot blurring tends to increase the absorption of light by the ink, the effective ink surface coverage is larger than the nominal one, i.e., dot blurring induces its own contribution to dot gain. Dot blurring can also be accurately modeled by an increased n-value of the classical Yule-Nielsen model.