Roger:

It seems to me that Mary also does not understand the ' "better" with regard to what? ' aspect / perspective.

For example, if I am an evaluator and I believe the observed rate is more representative without "aftershocK" clusters involved then I may choose A. The predictor may also choose A since this choice may make it look like he/she beat lower odds on a successful prediction.

OTOH, if I'm an evaluator with an agenda to scoff at predictors I may choose B since it makes it appear that quakes are more frequent and it's easier for a predictor to score a 'hit'.

The question as given has multiple answers depending on the context or who's point of view is being assumed.

Just my .02 worth.