In my own work, I often work with Dave Oppenheimer, who has a standard set of times and distances for each magnitude.

But the point of removing aftershocks is to effectively maintain enough independent data points. Aftershocks are fine data, it is just that if half a dataset is one aftershock sequence all at nearly the same time, the results will be bogus, to pick an extreme case.

So some common sense and experimentation may be the best route here, along the lines that you are suggesting.

John