To get any meaningful results, we'll need a decent sample size. Magnitude 7.0+ quakes in Japan are a fairly rare phenomenon, occurring 1-2 times per year. If we pull out a somewhat larger data set of 136 such quakes in the last 100 years (https://www.volcanodiscovery.com/earthquakes/japan/largest.h...), we'll notice that the quakes are fairly evenly distributed between the months.

The mean is (of course) 8.33%, and the standard deviation 1.91%. All months fall within 2 standard deviations from the mean (March is highest at 11.62%, followed by Feb at 11.06%, Jun 10.51% and Nov 9.76%). So, nothing particularly interesting to see here. Also, if we do the same "three-month window" analysis you did, the winner is February (with 30.68%, somewhat higher than the expected 25%).