The extension of DNA confined to nanochannels has been studied intensively and in detail. Yet quantitative comparisons between experiments and model calculations are difficult because most theoretical predictions involve undetermined prefactors, and because the model parameters (contour length, Kuhn length, effective width) are difficult to compute reliably, leading to substantial uncertainties. Here we use a recent asymptotically exact theory for the DNA extension in the "extended de Gennes regime" that allows us to compare experimental results with theory. For this purpose we performed new experiments, measuring the mean DNA extension and its standard deviation while varying the channel geometry, dye intercalation ratio, and ionic buffer strength. The experimental results agree very well with theory at high ionic strengths, indicating that the model parameters are reliable. At low ionic strengths the agreement is less good. We discuss possible reasons. Our approach allows, in principle, to measure the Kuhn length and effective width of a single DNA molecule and more generally of semiflexible polymers in solution.