I tend to disagree with you on the value of predicting small earthquakes.
The purpose (as I understand it) of predicting small events is to demonstrate
in a statistical sense that whatever method Don (or anyone else is using)
is statistically better than random. Once that is established (or not) the method
can be extended to see if it works on larger events.

It seems to me that it is better to test a method (especially in a public forum
such as this) on small events which bother no one and about which no one
will get excited. In this way, the method can be tested while at the same time
not producing the side effects which prediction of larger events (if taken
seriously) might cause.

That's my case for starting out with small earthquakes, and the reason I
think is a valid approach even if it is not terribly exciting.