The sequence that is 87.5% is that you will have at least 1 head. If you had 1 head on the first trial or the second that condition is already met before the third trial so the third trial can be heads or tails. The only sequence that guarantees the third trial is a head is if you eliminate the trials that already had a head. If you eliminate the first two trials because you know the third trial must be a head to satisfy the criterion of at least one head the sequence goes from 1- (1/2 x 1/2 x 1/2) to 1 - 1/2 = 50%. that is what we have been telling you.

You are quoting the instances of at least one head as influencing the likelihood of the third trial but you already know trial 1 and 2 outcomes. If one of them, was a head then you could flip a tails on the last toss and still be in that 87.5%. So to determine the likelihood that the third trial will be heads you need to already know the outcome of the first two trials so the likelihood of those trials are 1- (1 x 1 x 1/2) to 1 - 1/2 = 50%. Since you already know those trials were tails the likelihood they were tails is 100% or 1/1. You are a victim of a gamblers fallacy of misusing probability to get a false answer.