Alex P's answer touches upon an interesting philosophical angle to this question. Suppose there are two strategies of game play in a four player game. Strategy A will reliably give you "good performance" (whatever that means for the current game - say 100 points) if carried out well; strategy B, if carried out well, gives you "bad performance" in 60% of your games (say 60 points) and "exceptionally good performance" 40% of the games (say 110 points).
In this (fictional) game, strategy A gives you better performance on average, but it may give you a lower probability of ending up in first place: in an environment where your three fellow players also play A, the game will be a toss up and strategy A will win you 25% of games, all other factors being equal, whereas strategy B will win you 40% of games. If your three fellow players play B, then strategy A will win only if all three of them have bad luck - less than 22% probability - and strategy B will turn it into a toss up again, so 25% wins. In both cases, strategy B will give you a higher probability of ending up in first place than strategy A.
Thus we see that even a relatively small chance of a relatively small improvement of your score over the alternative strategy may be worth a large risk of a really bad score - if all you care about is ending up first. However, you may prefer doing relatively well compared to your peers for certain, to having a slightly higher chance of winning overall but also a higher chance of ending up dead last. In that case, strategy A is better.
How does this apply to your question? Even if we leave the social aspects of the game aside (where players are more likely to impede further success of already successful players), going for clumped-up bunches of resources has some of the properties of strategy B: if you can ensure a steady trickle of resources, you will do well, but if you gamble the same number of settlements on a few numbers and they happen to come up more often, you will do even better - at the risk of losing badly if the dice don't go your way. Thus my answer is: all else being equal, spread the numbers if your objective is to do relatively well; concentrate the numbers if your objective is to have the highest chance of winning, and you're willing to accept a high risk of ending up last. (The key phrase in that sentence is "all else being equal", of course - in practice this consideration should be only one of many arguments when weighing different courses of action against one another.)