A good DSLR or video camera typically records compressed video in
YUV format with
chroma subsampling in 4:2:2. This means for every 4 samples of luminance detail, chroma is sampled in 2x2 chunks. In addition, the sensor has 14 bits of depth before it applies profiles, so banding is pretty negligible.
The Mavic records compressed video in YUV 4:2:0 format; this drop of the last :2 to :0 means you have half as many samples of color information in the vertical direction. You may see more color artifacts and moire of color alternation (green turns to brown and back as sensor moves vertically very slowly). In addition, the sensor likely only has 10~12 useful bits of depth before it applies profiles, which is 1/16 to 1/4 of the amount of color resolution that a DSLR would have, so banding is more likely.
When you decide to flatten the data to use only the lower end of the numerical range, roundoff and quantization happens. Take a number like 129.25, and turn it into an integer (129); not much is lost. Take a number like 129.25, divide it by 16 as a dull profile might (8.078125), turn it into an integer (8), and multiply it back up by 16 in post-production (128); you can see a larger amount of information was lost. A smooth curve should have values like 128, 129, 130, 131, ... and so on. If you did the divide-then-multiply operation, your final results would step a bit and result in 128, 128, 128, 136, 136 and quantization happens. This is also why photographers stress you should "
Expose to the Right" end of the histogram.
It's a cell phone camera. Don't push it around too much. If you like one of the pre-graded colorized artistic profiles, use it. If you want a faithful color with a little pop, go with D Cinelike. Honestly for basic scenery, the default linear profile is just fine and you shouldn't stress over it so much.
