There's several problems here.
First off, so-called "Best-Seller lists" aren't empirical gauges of that. For example, the NY Times Best-Seller list was cited, but they honestly do some editorializing (i.e. not purely based on sales and make judgment calls on whether a book should be on the list or not; for example
http://antickmusings.blogspot.com/20...imes-list.html). That's why I tend to favor USA Today for best-sellers based on actual sales (for print) (their algorithm is on their site).
Second, at least in print publishing (I'll get back to this point with eBooks), publishers don't have live stats on how well their book is selling (if they did, one of the problems of book returns wouldn't be as difficult as it is). Now eBook publishing gives more immediate reports than print... but at the end of the day, that also depends on the retailer. Amazon and iTunes provides relatively quick stats (under 24 hours). Other retailers, not as immediate.
Third, for print books, while it is inaccurate (sample size, demographic, etc.; something all best-seller lists face, including my first point), Bookscan provides sales for print books and is often used by publishers. Why is Bookscan important? Because it's a third-party source. It's not the retailer. It's not the publisher. Their agenda is to provide the best data, not their own interests. So who is the third-party data gatherer for eBooks (and this becomes more difficult when one of the largest retailers refuses to disclose their data).
So yes, publishers can give third-party the data given to them by Amazon. And then it'll take them more time to gather the data given to them from other retailers. But also, on some level, you have to suspect data being given to you by publishers. Sometimes they've fudged the math to favor them or to hype up a product.
In the case of the NY Times eBook best-seller list, here's some data on how they gather their data:
http://www.libraryjournal.com/lj/hom...dy_to.html.csp