If you toss a coin, and the first two tosses were heads, that does not influence(reduce/increase) the probability of future tosses: it is still 50% or 1/2
That is correct, but that's not a comparable scenario to the one in the OP.
Try the following real world experiment:
Grab a container of pennies and randomly lay out 100 pairs (200 pennies). Pennies that show heads we'll call boys, pennies that show tails we'll call girls.
Remove all pairs that show both heads, since couples that have two boys don't count.
Count the total number of pairs that you now have out and compare that to the number of tail-tail (girl-girl) pairs you have. About 1/3 of the pairs will be girl-girl.
So, for all pairs that have at least one girl, only 1/3 of the time will both kids be girls.
Or as unreality likes to phrase it: For all pairs that have at least one girl, 1/3 of the time the other child will be a girl.