The point is to communicate that it's not a steep step change. In the original caffeine-only model, it's a smooth exponential decay function. There is still measurable caffeine in your blood N hours later, at levels approximately 1/(2^(N/5)) as high as the peak concentration.
In the new model, it's still smooth decay, but it's a compound exponential decay which is spread out over a longer time period, and close to linear for a while after that, before going on a longer exponential decline.
In the new model, it's still smooth decay, but it's a compound exponential decay which is spread out over a longer time period, and close to linear for a while after that, before going on a longer exponential decline.