I thought the g limit applied to terraformables too?
I just checked, and looks like I was wrong on the g limit: it's the same 0.4 g with terraformables as well. I could have sworn I remembered ones lower than that, but that just goes to show how your memory can fail you.
Rather, the cause for failure is that the moons' surface temperature ends up too low, while still being a TC. Then there can be a "wrong" atmospheric composition as well. (A small hint of ammonia, and bam, it won't be an ELW.)
Also, more TC moons of ELWs than I thought: currently, 2,026 such moons out of 28,536 moons of ELWs in total. That makes it pretty rare though. Added this to the categories as well.
As for ELM parents that are terraformable: 506, which is interesting too, as the most common parent type for them is HMCP anyway. Of course, the parent has to be heavy enough to have a natural ELW as its moon, and enough of an atmosphere to push past the atmospheric moon surface temperature bug.