You have to go beyond the first section.
His main points are that (a) it's hard to standardize on the larger parts and (b) there is no low hanging fruit.
Please don't imply people didn't read the article. It's boring, unnecessarily rude, against the guidelines [1], and at least in this case I did read the entire article and agree with that comment.
[1] https://news.ycombinator.com/newsguidelines.html
> (b) there is no low hanging fruit.
There is no low hanging fruit when you divide up the pricing the way he did, but transportation is part of nearly every category that he divided it up into. From the data he shows we don't have any reason to think that transportation is not a single substantial fraction of the house cost, when you some up all the different transportation costs between the different sections he uses.
To make an analogy to programming, transportation is like an allocator, and his data shows that execution time is spent evenly between 20 different functions. But all of those functions allocate, it's perfectly plausible the program is spending 50% of it's time in it's allocator and that speeding up that allocator by 50% would reduce the overall execution time by 25%.