The PDF format is designed to be a container for page objects which can be added, deleted or reordered very easily.
By very crude analogy, you can easily add another Paragraph to an HTML document by adding a set of <p>
tags around some new text data. You can remove the paragraph by deleting the tags and text. You can re-order the paragraph by moving its order relative to other paragraphs.
So, dragging pages from one PDF document into another simply adds the Page object data from the source document to the destination document. So there is no difference between 'really merged' and 'acting as a container for more page data'. Page data is indexed, so the PDF document knows how many pages it contains and in what order they come.
Similarly, deleting a page removes that Page Object from the PDF data stream and re-indexes the pages. You can test this by deleting a page and checking the file size reduction.
Obviously, Apple tries to conform to the PDF specification (otherwise they wouldn't be PDFs, they would be 'Apple files'.) Any divergence is a bug. However, the same page information can be written in a variety of ways: different encoding structures, colorspaces, additional metadata, etc, so it's possible for Adobe's PDFs and Apple's PDFs (and others) to be different sizes and 'styles' within the PDF spec.
Mountain Lion
this method: open PDF doc2, open PDF doc1, drag thumbnail from doc1 onto doc2. The doc2 thumbnails display pages from doc1 + doc2. But once saved this new doc2 does only contain doc2 pages. – dan Dec 10 '14 at 14:32