Strictly speaking that shouldn’t be possible in Scrivener, but there may be some other thing going on that makes it look like that is what happened.
As you probably know, ebooks are formatted using simple HTML. If you enable the source file export setting in the General Options compile tab, you’ll get the files used to create the .mobi file. You can then much more easily see what is actually happening—which file is the hyperlink loading, and is it using an anchor point found within that file, etc.
Not sure what you mean about the “actual ebook”, though. The way it shows up and functions in Kindle Previewer is going to be what everyone else gets.