I can’t figure out why the DLT linear model has a very slight error between the actual level and the calculated level I derived from the underlying math.

Specifically, when I attempt to derive g(t) / D(t) to validate the level calculation, the slope component does not reconcile with the model’s own level output. It almost does, but it’s consistently off by around 0.1% - 0.4%.

Is there something I’m missing? Did a degree of freedom get burned somewhere or something of that nature?