Having 2 each of BM, GM, etc. is implicitly labeling 1BM, 2BM, 1GM, 2GM, etc. If you sampled with replacement, you have to admit the possibility that you draw the same BT twice. It's a stretch to suggest that the statement is ambiguous in this way, as it would imply that the father could have the same child twice.
Trying to figure out why it seems wrong at first: it seems my first reaction would be emotional, that the odds of having a boy should be 0.5. However it is exactly because the total probability of a boy has to be .5 that the probability of a second boy has to be less than 1/2. The fact that you can pair "boy" with any enumerable attribute that can bring the probability down to 1/3 is ... funny.
