I couldn't find the house at 72 feet elevation listed.
I must've missed where that was posted; that's the house at the end of Ice Pond Lane, at 43.866210, -68.887392, with nice tall trees all around it?
Here's my final figures for this pic then, which will hopefully be useful to
@eenor for when he does his tests next year:
Elevation of lawn above mean sea level: 16 feet
Elevation of camera above lawn: ~5.5 feet
Distance to Marblehead Island: 2.56 miles (b)
Elevation of summit of Marblehead: 33 feet amsl
Distance to high point of NE portion of Matinicus (as seen to left of Marblehead):: 16.61 miles (y)
High point of NE portion of Matinicus: ~105 feet amsl (including estimated 35 feet tree height)
Level of water: 5.3 feet below mean sea level
Camera height above water: 26.8 feet
Summit of Marblehead above water: 38.3 feet (a)
Summit of NE Matinicus above water: ~110 feet
Predicted hidden amount of Marblehead (given standard refraction): None
Predicted amount of Matinicus hidden: 54.5 feet
Predicted amount of Matinicus visible (including tree height): ~55 feet (x)
Size in picture of Marblehead: 35 pixels
Size in picture of Matinicus: 9 pixels
Relative size in picture: 3.9x
Predicted relative size in reality: 4.5x (using (a/b)/(x/y))
Predicted amount of Matinicus visible using relative sizes: 63.9 feet (using 9ay/35b)
So it actually appears that there isn't that much extra refraction happening, as far as the highest point of the island is concerned: indeed, if the trees at that highest point are 43 feet tall, the image matches perfectly with what is predicted for standard refraction. And if they're 35 feet tall, it only requires refraction to be about +10% of standard for that to fit - well within 'normal range'.