Looks like Safari (webkit? Dunno) "Copy Image" gets a lot more than just the image data.
To see all the types data you get, run osascript -e 'clipboard info'
in a terminal after clicking "Copy Image" (make sure to copy the image after pasting the command of course).
I get something like this (not sure what it all means exactly...) with Orion:
TIFF picture, 861652, URL, 23, «class urln», 10, «class utf8», 23, «class rtfd», 85280, «class weba», 71263, «class HTML», 2488, «class 8BPS», 858567, GIF picture, 188686, «class jp2 », 60541, JPEG picture, 94824, «class PNGf», 508042, «class BMP », 859214, «class TPIC», 844120, «class ut16», 48, string, 23, Unicode text, 46
With FF:
TIFF picture, 3002406, «class HTML», 238, «class 8BPS», 4539052, GIF picture, 600518, «class jp2 », 152409, JPEG picture, 257237, «class PNGf», 1603864, «class BMP », 3000054, «class TPIC», 2949398
This utility gives you more human-readable access to the data: https://langui.net/clipboard-viewer/
I think the intention on macOS is that the application which does the pasting is responsible for choosing from the types of data in the clipboard.
For example, if I paste an image (which was a link also) from Twitter, in an image editor I get the image, but in a text editor I get the link URL (not image URL).
I'm not sure what is preferable - the aforementioned broad copy, or a narrower, explicit copy.