I recently acquired a Kiribati 5 cent coin dated 1979.
The context of said acquisition (as part of a coin set sold for a bit under $4 that was supposed to be overpriced... Kiribati, Zambia, Mozambique, Cabo Verde, Bahamas; even if all the five types were common - they were - who wouldn't buy that for $4?) made me suspect that the coin might have a higher mintage than Numista (and apparently Krause) says; I've since received more definite proof of it (the type appeared in an issue of a Russian coin collecting magazine - even though there were 120,000 copies made of that issue, and the official mintage of the type is much less than that).
However, while trying to find previous discussions of this type on Numista, I stumbled on this thread. I checked, and sure enough, my example was magnetic.
I would've bumped the old thread, but it wasn't in the right forum (Numista catalog), so I started this one instead.
Best I can figure out, for a few years starting in 1992, the KM 3a type (same as KM 3 except magnetic) was produced by Kiribati - with a frozen date of 1979.
So there aren't actually any 1992 dated Kiribati 5 cent coins (regardless of what Krause says), but the KM 3a type would indeed have been made in 1992 (or a bit later).
Note that the mintage listed for 1979 only applies to the non-magnetic type, and that the listed pic of the "1992" magnetic type actually says 1979 (as does the legend). [In fact it appears to be the exact same pic as for the 1979 type, complete with small spots, scratches and dust specks in the exact same places.]
However, judging by the rarity scale, most Numista members that had one of the magnetic 1979 coins would list it under the "non-magnetic" 1979 type, presumably because their coins were not dated 1992. So a straight fix probably wouldn't work very well.
What should be done about this? Something certainly should (I'm not going to enter my magnetic 1979 5 cents under either of the two types), but I'm not entirely sure what could work (perhaps creating a separate third page for the magnetic 1979 is now the only semi-reasonable possibility).