This is their key to excessive warmth prediction: "The scaling factor β dictates the size of the coastal-inland pattern. Using our dynamically downscaled output, we compute that α ranges between -0.10°C (summer) and 0.15°C (fall). In contrast, β varies little by season, ranging from 0.26 (winter) to 0.39 (spring)."
Their prose is the skilled prose of professional liars. One might think that α and β are somehow calculated from weather patterns (present or even the unknowable future). But they are not. They are simply a garbage side effects of their statistical model which is filled with their made-up assumptions. They cannot possibly know whether there will be more land breezes or ocean breezes in the future.
Yep. At first the models probably predicted some cooling. After tweaking the variables enough the desired results were achieved.