Strictly, I think the proper name since standardisation is G.998.4, while G.INP was the provisional name during the standardisation process (similar to G.vector and G.993.5). While Broadcom came up with the PhyR idea, and had their own product based on it, I'd guess that standardisation has changed the implementation somewhat.
Also note that it is technically feasible for FEC to be turned on without interleaving - resulting in D=1, but for RScorr and RSuncorr to accumulate - with nothing to do with G.INP. If the RScorr and RSuncorr are accumulating due to the FEC process alone, you'd expect to see D=1 and R>0. It isn't going to be as effective as when interleaving is also activated, but it still works. I've seen it happen plenty of times on the upstream, but (IIRC) only once downstream (and if I am wrong, then it is zero times).
It'd be interesting to see what the values are for "INP" (and INPrein, if it is shown), "delay" and "R" when the "FEC errors without interleaving" scenario is happening downstream.