His results are very general. They depend only only on the fact that keywords with more words have lower search volume. For any industry for which this is true (i.e. any), you will find qualitatively similar results.

If you think about it, it's a trivial result. In any other industry if I tell you "I'm going to focus on increasing the ROI of 10% of our sales", you'd laugh. Why not focus on increasing the ROI of the other 90% of the sales?

However, it does depend on the account. On an account where you have the same number of 2 word keywords as 5 word keywords, of course the 5 word keywords will have a tiny tiny percentage of the impressions (say 1%). But obviously a marketer who goes for the long term strategy will tell you that you need 100x more 5 word keywords, in which case the 5 word keywords would have the same number of impressions as the 2 word keywords. But in comes the author's argument that you can do virtually nothing to manage that many keywords (we'd be talking millions here).