May I suggest that you begin to increase your expected lower magnitude if you
expect a reasonable sized quake, then call it for a reasonable size. The evaluation
will test from there what your probabilities are. Check and see how often events
of Ml 3-4 occur in the area you had predicted for and decide is this an every
day occurrence or is there something unusual about it that would cause you to
expect a larger magnitude than you might normally expect.
This would eliminate those everyday quakes and start to pinpoint the regionally
significant ones, which is what you really want to do, I assume.