While it's not statistically valid evidence, all 4 people in my office who owned an Xbox had to have theirs replaced. One of them had to do it twice. That's hard to reconcile with a 5% error rate.
Presumably the last Xbox people get actually works correctly. By that measure, my office had 5 occurences of the problem and ended up with 4 working consoles out of the 9 total. That's starting to sound like the 50% failure rate.
You could argue the failure rate of the console isn't the best measure for the situation; maybe a better measure would be how many times customers ran into the issue. My office averages 1.25 failures per Xbox owner, and 100% of the owners had the problem at least once. That's not good.