Pretty sure they didn't come up with the same image. Images 1, 3, and 4 are the three images the GP generated and they put the Imagen-generated image (2) into the set for ease of comparison.