I don't know, I'd say, about this much.
First of all, you were super imprecise with your "outline" of the lower moon.
1. Use a thinner line, so you can be more precise.
2. Don't fill it in with a green color. The filled in color just hides the lack of precision.
Second of all, you chose a moon that was partially behind the horizon. A significant chunk of the bottom is being obscured. And yet you still outlined it with an ellipse...
I cut out the bottom moon and overlayed it directly over the top one. This shows how imprecise you were.
Notice several discrepancies between the overlay and the moon below it:
1. Bottom: The moon is partially obscured by the horizon, obviously.
2. Top right: The moon is partially obscured by the moon above it, obviously.
3. Bottom right: Here is where the shape does actually change slightly. This is an expected result of refraction very near to the horizon.
Or, if you prefer a table of measurements:
I labelled the moons 1 (bottom) to 8 (top). If there is a dash (-), it means that diameter was partially obscured, so I couldn't measure it.
moon horizontal diameter vertical diameter
1 - -
2 59 pixels -
3 59 pixels 53 pixels
4 59 pixels 55 pixels
5 60 pixels 55 pixels
6 59 pixels 56 pixels
7 59 pixels 55 pixels
8 59 pixels -
Remarkably consistent! The only outlier is #3. It's vertical diameter is
slightly less than the others. This is consistent with what we expect from refraction: the size in the vertical direction will appear slightly smaller. (i.e. it will appear slightly squashed). #2 would also probably have a smaller vertical diameter, but since it is partially behind the horizon, we can't precisely measure it.