Just select your dataset yourself and throw out the statistical tests compared to the measured data, seem to be practice on both sides of climate "science". Until they stop messing around with statistics I'm not going to believe either side, got better things to do like restarting my CFC production plant!
But to be serious for a minute here, climate science really annoys me at times. Many times you e-mail authors, and I'm talking about both sides, to request more information about their datasets they either say it's confidential (really???), lost, destroyed, ... or they don't respond at all. Those that do have either an inconclusive end result or questionable practices. If you cherry pick your data to lead to your result it's not difficult to come up with the conclusion you want, combined with the staggering lack of statistical background knowledge. It's one thing to remove noise from your data, almost all researchers do that when they publish. If you're measuring over a few months or years you're bound to have a few extreme values that aren't representative. But then you should also mention what you did, why, and what the influence on the overall dataset was.If you go further then you should execute the right statistical tests to verify if the chosen samples are representative for the entire dataset. And until they properly do that I'm not taking either side serious.