At any point in time, k, the genetic makeup of the population is an acculation of the results from the past .... which will include deceased individuals. So maybe I changed tack a bit there, but it's been a while since I even looked at what I was saying here.
Well, that may explain why you aren't making sense. If you don't understand the position you are trying to defend how can you make sensible arguments?
The point you are meant to be explaining is why the effect of selection on the distribution of genes in the current population should be a function of the distribution of genes in the current population, rather than the previous generation. And yet in the case of drift you DO use the previous generation.
Interestingly, this means that if there were no drift, the previous generation would have no relevance to your formula.