It seems to me that it's only "the conditional probability that the recipient of altruism contains an altruism-causing gene given that the altruist does", if the altruist is the only source of that gene. But I just pointed out that in real life we have no reason to assume that the altruist is the only source. So when a successful gene spreads through the population, the 1/2 and 1/4 stuff should no longer be relevant. It only holds for a newly introduced gene, right ...?

As I said, r is the limit as gene frequency tends to 0 of the conditional probability, so as it becomes more common the true conditional probability rises. Remember, however, the gene for altruism has to satisfy rB-C > 0 for all periods in which it becomes more common, including the initial brand-new period. As time passes, the "true" r will exceed that of these calculations, making the "true" rB-C greater, so it will still be positive. But because the gene will only get to that era if initially rB-C>0, the altruism policy it enacts has to be no more altruistic than the pessimistic estimate of the conditional probability warrants.