Thankfully I made it a little easier to look up this sort of thing in the database, with reference keys for parent relationships etc. Probably not 100% perfect, but good for estimating things like this. It looks like this method of querying it agrees with what Marx reported above, except for the overall ELM number:
Right now it looks like out of 200,305 ELWs, 2885 of which are ELMs detectable by proc-gen name (2476 of which have detectable parents in the data), 137 of them orbit a gas giant as a direct parent.
The ELM overall count is probably missing cases where the names don't match up by string comparisons, or due to other flaws in the data. Sometimes EDSM itself can figure out things that aren't then passed down in a meaningful way through the JSON dumps.
