without being familiar with your software, I suspect the issue is the inability to estimate something which isn't present (sic). For moving water with sun reflecting off the surface, you get intense highlights spots that really cannot be recovered. One image a hotspot is present, next it is gone. Poor software has to make educated guesses as to what really should be there, which of course is almost impossible with a rippled water surface in which reflections occur at very high spatial frequencies. What one could hope for is removal of the chromatic fringes around these hotspots. Do note that the colour fringes aren't necessarily caused by defects in the lens, as the water also adds refraction components to the image.
Assuming a tripod is used, take two exposures one pulled down to keep most of the highlights with detail, another for the surrounding landscape. Remove the purpe fringing from #1, brighten it slightly then add the water parts to #2. My hunch is that this will always be an improvement over automated procedures.
Or you could try out PTGui, which sometimes works pure magic. Other times, it fails miserably.