PDF is just a wrapper that holds data. It can hold all sorts of data. Not all programs write the same data, and not all programs read the same data. Some programs write redundant sets of data in different ways. Some programs can read all the redundant data, and some just read portions. AI and Inkscape do things differently, and that is often frustrating.
AI (as I understand it), will essentially save two versions of your file into a PDF when you select “Preserve Illustrator editing capabilities” in the PDF save dialogue. AFAIK it does not do this if you “print to pdf” from the print dialogue, but it does RIP the appearance data. That could possibly be a cause for a PDF/X version of a shape to be slightly off-dimension from the vector inkscape version.
PDF/X is for sending to a printer, and leaves out a bunch of info that a printer does not need/want (info that a CAM/CAD process might be interested in seeing).
A Layered PDF might not be as good to send to a printer (who knows what layer they would end up printing, maybe all layers at once), but is perfect for giving to another designer who will continue to work on it, or for a file that is being used for another type of process. I use layered files with my plotter to differentiate between colors (load a blue roll of vinyl, select the blue layer to send to the plotter, etc)
Someone who enjoys coding and poking through code might do us a favor and try looking at the differences in the code of several versions of a simple file (say a 10mm square with a 2mm stroke?) that was created with “Save As… PDF”, “Export for screens > PDF”, and “Print > to PDF” (in AI, and also whatever the corresponding options are in Inkscape).