If I understood correctly, you're taking the number of quakes and the number of events, and dividing one by the other to find the expected interval between events. As Mike points out, that only works if events do not tend to come in clusters.

Another choice is the median time between events, which will perform even worse (underpredicting the intervals between events) if most events are in clusters, because usually the previous event will have been in the same cluster, and the long intervals would carry less weight with medians.

One way to get the right answer is to count the number of days since the last event for each day since the catalog became complete, and then do the statistics on the set of counts.

As an example of the difference, if in New Madrid, an entire 100-year catalog consisted of 100 events on the same day in the first year, your method would suggest one event per year, or an average time since the last event of half a year. My method would find time-since-last-event ranging evenly from 0 to 99 years, or an average time of 50 years.