I hate coming to pers defense, but I think this is just an issue of a useful stat being improperly used. (Az sel + iz sel)/(Iz elig) just compares the size of the block of people who will promote in the fy to the size of the block of people who's date of rank meant that they should pin in the fy. I'm not a pers guy, but that actually does seem like a pretty useful ratio.
Once you get into really small sample sizes like o5 selres nfo females, I suspect you'll always get weird artifacts like 250%. If you only have two IZ eligible, it doesn't take a lot of az eligible to reduce the usefulness of the stat. Or it could be nefarious intent by pers to lie to the masses.
What number would you prefer pers publish?