I don't understand the logic behind working on the jpg instead of directly on the NEF. LR edits are non-destructive and amount to nothing more than a set of instructions stored in the catalog and used when you export an RGB image (tiff, psd, jpg). You can edit a NEF from now until next year, revising and resetting the edits endlessly or making multiple versions as Virtual Copies, and the NEF itself will never change. Your workflow seems unnecessarily complex and clumsy; you are not protecting the NEF with it.
Thanks for the feedback, I am pretty familiar with LR's nondestructive capabilities. I am not trying to "protect" the nef - it is actually easier for me to edit only one of the files and then sync the changes because they need to be exactly the same. It saves time rather than doing the same exact edits twice in a row. If I needed the images to look different, I would spend more time editing both the jpg and the nef as well. As it stands, I need the two files to mirror each other precisely. I could edit the nef and then sync the changes to the jpg, but the same problem arises.
So why not perform the edits (cropping and whatever else you choose to do) on the NEF and then export the result as a JPG? Seems to me this achieves your goal of having both an NEF and a JPG that should be identical (except for JPG compression which is unavoidable and probably not noticeable).
By the way, I hope you realize that when you crop an NEF or any RAW or DNG, the original image isn't cropped, and there's really no way to produce a new NEF file that is cropped and contains fewer pixels.
What dj wrote regarding crops is equally true for all and any edits you do to the NEF, including any you paste from the jpg. Those edits exist only in the catalog and additionally in an XMP file if you opt for that. Thus the modifications to the NEF will be viewable only in your LR catalog or by somebody else (with the accompanying XMP) in Adobe software.
Also, you should consider that color-critical edits cannot simply be pasted from a gamma corrected image rendered in a narrow RGB space to linear and wide gamut image data and be the same. If it's a jpg from the camera it is in either sRGB or Adobe RGB, narrower gamuts than LR's linear ProPhoto RGB and colors have already been changed in the camera by both the Rendering Intent and the Tonal Response Curve. Moreover, the camera's jpg has been built on a Nikon camera profile while both viewing within LR and exports from it are on the basis of an Adobe profile that may seek to emulate the maker's profile but will never be the same.
You say you do WB adjustments to the jpg. How are +/- tweaks in a limited range applied on top of the camera's WB supposed to translate to WB for the NEF when even the UI (Temperature and Tint) is different?
I have had many years experience photographing paintings for show catalogs and your method seems fundamentally flawed to me.
I am photographing them for "before" pictures for art conservation. I am adjusting the WB only to a color bar that has been placed next to the art for that purpose; it is just to make sure everything is consistent from the before photos to the after photos. Some clients want the raw files and others want the jpgs, so we edit both and export the nefs as tifs since, as you said, nefs don't export without the xmp file. Tifs usually work for our clients that want raw images because they are still lossless. The system works for us; I am not really looking to change that although I truly welcome any feedback you have. I am really just trying to figure out why the Sync function is failing when it has been working with this method before.