A planet 34 AU in diameter would collapse under the weight of its own gravity. The 17 AU would be the “beamwidth”, think of it as a bell curve, containing about half the energy, the rest would be smeared out, so a 17 AU planet would just fill about one (well matched) pixel.
I blew up the photo as big as I could and the planet just spans about 4 mm on my screen. Then I measured from the planet to star, about 152 mm. The planet is 660 AU from the star. 660 AU / 152 mm = 4.38 AU per mm, so 4.38 x 4 = 17.4 AU. The planet spans about 17 AU on the scale of that photograph. That does not mean that that planet is 17 AU across, it means that the optical resolution of the camera taking the picture is about 17 AU at that distance. It cannot distinguish a 1000 km diameter planet from a 1,000,000 km diameter planet.
The number of pixels used in displaying an unresolved target in a JPEG has nothing to do with the target’s actual diameter. The enormous size of the star itself is also almost certainly an artifact. It is how large the image of the star bloomed during the long exposure. The far tails of the diffraction bell curve, not the size of the star.
I must admit that I was surprised that the prospective discovery of a PLANET that large was not creating a bit more excitement. But your evaluation makes it clear that saying much more than “we can see directly that some sizeable object is at this location” exceeds what can be stated with certainty.
It is still a remarkable achievement.
Thanks for taking the time and applying your expertise to tease out these details from the image.