RHESSI has data for over 19,000 flares. When Ed and I first started mapping with MEM_NJIT, I just randomly picked well-resolved (in the Clean images) flares.
Most of the results are from when the program was first being tested on RHESSI flares. You'll notice many of the MEM_NJIT maps have "pointy" contours, especially compared to the the other algorithms -- this is an effect that was eventually fixed by adjusting the spatial scale of the map. For more recent results (i.e. what we plan on publishing), take a look at our images. You can also view the graphical comparisons by checking out the plots page.