If we were not able to identify a small number of relevant patterns in a large sea of options, how would we learn anything?
If you've got a mountain of data and are willing to run a lot of models you can indeed identify ones with good tradeoff between prediction and generalization errors. It's very possible that some of your models are weaker than you expect due to cherry picking, but it's even more possible you've identified some useful trends.
As you said, he seems to be restating very common knowledge about economies of scale. Since he's not saying anything terribly surprising, I don't think there's much need to get up in arms about multiple comparisons.
Moreover he's looking for the on-average rates of growth of these economies of scale and identifying outliers which will likely be interesting. I think it's pretty exciting research.
There's an enormous difference between saying "I think X is true," and "I have significant evidence based on observation and theory to support the conclusion that X is true."
Also: http://lesswrong.com/lw/im/hindsight_devalues_science/
There's a section in The Black Swan devoted to the fallacy that we tend to project the fact that we understand the rather narrow field of physics very well - to the extend where we can predict things in that domain reliably - to an expectation that we can do the same for other fields, when in fact, it's physics that's the outlier in our ability to understand it so well.
In the end all his efforts seems to come down to the conclusion that cities exists because they provide economies of scale. I couldn't provide statistical proof for that fact to save my life, but still .. well, duh?