Just to be sure, I'm more used to seeing 50% kill fees, at least from Asia and West Africa (for Western or Western-run publications). 50% is customary, and with established clients, unsaid until the time comes
It certainly depends a lot on the type of publication (and the nature of the article).
The treatment of articles upon which publishers "choose not to proceed" is very important knowledge going in. Perusing the whopayswriters JSON feed* there is a wide variety of behaviour described, not all of it making everyone look good.
* In the absence of full-text search, left as an exercise to the reader
