Not your standard power of 2 question

You can read a lot about power of 2 textures and not power of 2 textures. A lot of the data is pretty dated and none of the best solutions presented for 2D (specifically UGUI) apply in my case. Here is our situation:
We have a mobile game with lots and lots of large images for character profile pages. We use raw image to show these. There is no sprite sheet option here because were talking about images in the 800x800 range.

Here are the knowns
Setting our texture type to normal has the smallest listed file size in unity
Setting our texture type to “to largest” increases file size and is bigger even when quality is set to low indicating the compression doesn’t help

Whats not clear: In this day and age, what is being done to NONE setting images.

Pros

  • They use less space according to the
    preview window
  • They correctly report
    the image size via code texture calls
  • They are not distorted when you use
    them in a raw image

Cons

  • All the cons are from a while back and its not clear if they still apply.
  • Its mentioned they use more memory at runtime. Is this because they have to be copied into a new texture and so at some point there are 2 copies until garbage collection happens?
  • If they do have to be copied into a power of 2 texture at runtime, then there is likely a one time performance hit when the texture is shown

If indeed the cons are real, what is the best practice for something like this where you cannot use a sprite atlas. Do you pad the image with black borders and just suck up the worse file size in order to have better performance?

Does anyone have any sources confirming the cons I listed are still true?
Does anyone have any other pros\cons not mentioned here?

Thanks in advance for any feedback people may be able to provide!

Lot of people are following this question. Unity people it would be great to get some official answer. Almost 200 inquiring minds want to know!