My suspicion of the 50% figure is based on my sense that while maybe 50% of all marriages fail, not 50% of all married people end up divorcing. If I'm not mistaken, second marriages have a much higher failure rate. You see people with two or three divorces, and others with one marriage. I think perhaps the confusion is in number of divorces vs. number of people who have been divorced. For every 42 year marriage, you have someone thinking "fourth time's the charm!"
the reason it gets screwed up, no matter what the percentage, is people don't understand what is being counted. And you're right, second marriage divorce rates are higher than first marriages- that's what skews the sample.
Where they get the % of divorces from is as follows:
Every year, they find out the number of marriages.
Then they take the number of divorces.
If 1 million marriages took place last year, and 400,000 divorces took place last year, they'll say (very unscientifically) that 40% of all marriages end in divorce.
This fails to take into account that in addition to the 1million marriages last year, there were, say already 50 million married couples. A huge, uncounted pool of marriages that they continue to ignore.
Of course, if they reported the actual % of ALL marriages that end in divorce, it would make it seem less 'mainstream' (gee, I thought EVERYONE was doing it) and that would cut into the divorce industry's profits.