This doesn't prove what Amazon wants to claim it does.
1) Authors don't just earn royalties from ebook sales. They also earn royalties from hardcover/paperback sales.. and if dropping the ebook price causes lowered paperback sales, then the authors could earn less.
So merely showing that lower ebook prices increase ebook sales does not prove Amazon's claim that lower prices increase authors' royalties.
2) Amazon's test may not even prove that it would increase ebook sales. Hachette earns money from the sale if the book is sold at BN, Amazon, or any retailer. If only Amazon lowers it's price to $10, then Amazon may just be getting additional sales that would have gone to other retailers. So the only thing Amazon's test proves is that Amazon would be better off in a fictional world where Amazon has the lowest price in the market -- 33% lower than any other retailer.
You're right though, that it doesn't completely prove the point, but I think it definitely makes a point. People who wouldn't buy the e-book at 14.99 are probably unlikely to go out and buy the same book from another retailer for 14.99. The question is if it really increases the market share of the book, or if instead lowering the price on amazon makes everyone who would buy the book would just go to amazon for the cheaper price. (Given that e-books cost nearly nothing to distribute, I would think that the price drop would pretty easily pay for itself though)
Amazon has put forth a number that is an aggregate.. it takes into account your value judgement, mine, and 100,000 other peoples judgement of the value of an ebook.
Which reminds of a 3rd case that could affect Amazon's test: people do not purchase a generic ebook. They purchase a specific ebook, and the value of that ebook very much depends on the content. A poorly written book is worth less than a well written book.. and a book on a popular subject is worth more than a book on an unpopular subject.
If the book (or books) Amazon tested were more mispriced than the typical book (so they showed a larger increase in sales because $10 was closer to the actual value of the book.. whereas maybe $12 (or $20) represents the value of the average book).. then when all books are lowered in price, the result may be a decline for Hachett and the authors.
So I've put forth 3 cases that may affect Amazon's test. If just 2 in 13 people are affected by ANY of these 3 cases, then Amazon's conclusions are wrong, and the net result would be a decrease in revenue.