With the caveat that I’m spit-balling ideas here, my thought on the out years goes like this.
When you think about GI sales, I think it generally consists of a group of initial “anti-selectors” that effect early year mortality and (a hopefully larger group of ) reasonable risks. I think in the early years the anti-selection overwhelms any possible mortality improvement. If less-than-healthy people have the opportunity to anti-select, they most likely will (which is why marketing/distribution is so important). But the further out you go, you are probably left with (or at least approach) the relatively “normal” risks, maybe somewhat less than population mortality.
So my thought is that they are probably reasonably subject somewhat to population mortality improvement. So not additional improvement, but just population improvement. Not sure how much the mortality in the out years has on profitability though without modeling it. A lot of the inforce is gone by then.
You also want to consider your renewal lapse rate experience. High lapse rates implies more anti-selection.