Just look at the data you have here:

John Hand Brown Cindy Jones Thomas More
[download]

Knowing nothing else about your data, I look at this and decide that there are three seperate names here, which are "John Hand Brown", "Cindy Jones", and "Thomas More". But it's quite possible that the names are actually "John Hand", "Brown Cindy", and "Jones Thomas More", or perhaps some other combination. Consider that the human brain is much, much better at solving ambiguity than computers are (or at least making a solution that is closer to reality).

If the data above is repesentative of what you have, then I don't think you're going to find a solution with even an acceptable failure rate.