I also don't use fuzzmeasure, so I don't have anything to add on the software setup.
However, are you sure there is no low freq AGC (auto gain control, compressor?) somewhere in the input chain? What mic, pre-amp/soundcard set up are you using?
The only other thing I can think of is some low level external noise that may have been picked up ie other than the test sweep. Have you tried the same set up in another room? If you get the same LF late onset increase in level then there is probably something wrong with your set up?
I believe it was user error: the gain on the loopback channel (ie, channel 2 output into channel 2 input; used for measuring the internal delay and FR of my interface and compensating accordingly was set too low). Now it's set appropriately, I'm getting consistent decay in the sub 100Hz region. Must have been that.
For the vast majority of everyday operation, Loopback correction is not necessary. If you need extra S/N ratio, use longer sweeps. That is just as good as multiple ones combined. But lose the Loop.
If you see those odd cut off blurps at the front of the graph, i.e. end of decay, make the IR window longer, as long as the graph axis if necessary. That will clear it up.