Have you ever made a best fit curve in your life? We are dealing with a very simple exponential curve. Just because the actual growth rates fluctuate from month to month does not change they fact that M3 grows exponentially and there is a constant rate factor that plugs into the equation that nicely tracks the above graph. When you make a best fit curve, the object is not to hit every point. The best fit curve will have some portions that are above the actual data and some portions below the actual data. This really is not that complex.
It's pretty easy to understand why fission is a simple exponential process or yeast growth is exponential.
But I don't seem a similiar case for either GDP or money supply or even human population.
No *you* are reducing a very complex phenomenon to a very simple exponential curve.
The data is far more interesting than your model of the data (which has absolutely no predictive value).