Glen on a “A sprinkle of Cocoa” has summed up the situation very nicely,
“on the Mac we used to have this feature which I will call Round Trip Editing, wherein a user could make a drawing in one application, copy and paste the drawing into another application, and then later copy and paste back into the original application, and still be able to edit the drawing. Generations of Mac users relied on this feature to go from applications like ChemDraw into PowerPoint and back again. This feature has been lost as applications transition from using the archaic PICT clipboard flavor to the modern and beautiful PDF clipboard flavor. There was no direct way to embed large data in Apple generated PDFs, and thus developers were left on their own to munge the format if they dared. And, with no standard or expectation of data embedding, applications did not bother to preserve the original PDF resulting in data loss”
However it looks like the situation has now been resolved Starting with Mac OS X 10.7, developers will be able to embed arbitrary XML data in the PDFs they can generate with Apple's APIs. This means that in the case of chemical structures vendors can embed structural information in the meta data that can be then used to generate the chemical structure when copied back into the chemical structure drawing package.
The additional metadata may also have added benefits, it might be possible to use the data to allow substructure searching of documents. The inclusion of SMILES or InChi would probably be the most compact way of doing this in a similar manner to the way I attached data to image files.
This of course will not be limited to chemical structures, any PDF object can now have associated metadata.