As written in the comments (see http://rawtherapee.com/forum/viewtopic.php?f=3&t=5406) the issue is not intrinsic to 12 bits, but it is caused by a wrong interpretation of the RAW data by dcraw (the common RAW decoder used in most Linux software). The link above provides a temporary solution (set black point to -450) while waiting for an updated dcraw (you may try to compile the latest version, too). Thumbnails won't be correctly displayed anyway.

However, don't jump immediately to 14 bits unless you know you need them: I used 12 bit for years and the difference with 14 bits is barely visible.

Not only is barely visible, often the different is actually nothing: if the camera has about 13 bits of original dynamic range (the D5300 for example http://sensorgen.info/NikonD5300.html) and you shoot at 400 ISO, you already have less than 12 bit in the input data. 14 bits are not needed beyond ISO 200-400 ISO. Surprise!

If you use ISO of 100-200, you can still check here for real examples to understand what you actually gain for those two sensitivities: http://www.diyphotography.net/12bit-vs-14bit-raw-and-compressed-vs-uncompressed-does-it-matter/

You may want to check what is the file size difference for photos with extreme dynamic range at 100, 400, 1600 ISO. As for extreme dynamic range, stay inside a room of your apartment, one with some dark shadows, while at the same time you have a sunny landscape (buildings, sky, and so on) visible through the window. This should offer you more than 14 bits of dynamic range. Set the camera to Auto, check the exposure, then set it to manual with that exposure. Use a tripod, take one shot at each of the sensitivities I listed, then change bit depth, take again the same three shots. Download the files, compare the size and check if the additional space 14 bits require is justified by the additional highlights and shadows you can recover.