You lowered the gain on the H4n until it showed no clipping, while continually recording? So what we are seeing is the result of the initially clipped recording, and what happened subsequently while the gain was lowered?

Then it certainly appears that the gain padding happens PRE-A/D. Good news.