EIther that, or there is a kind of 'flag' like the rotation flag. The picture isn't actually rotated by the camera, it just stores that this photo was taken like that. Consumer photo software rotate right away based on the flag and to the user the picture seems to have actually more pixels on the side (portrait). Highend software however usually ask before rotating.

Might be that a certain WB estimate is stored in the raw file by the camera so that the PC software has a nice starting point (but remains free to ignore), something rawcam doesn't do (I don't know if dng even supports that). The situation with jpeg is different. The color balance is "baked" into the image, meaning that the pixels are actually that color.