Yes, this has nothing to do with OCR. What Acrobat is doing is optimizing the document by segmenting the scanned image into different components using different compression algorithms for several sections.
Images in PDF are rasters compressed by one of the following;
en.wikipedia.org
If you open the document in a hexeditor (or vi like I did), you can see the different XObject compression schemes used, I found;
* FlateDecode
* JBIG2Decode (only one raster uses this)
* DCTDecode
If the conspiracy theorist had half a brain, they would have guessed this had to do with a compression scheme since the images are divided into subdivided square areas.
BTW these are XObjects, Illustrator is intepreting them as layers but it really has nothing to do with layers.