By pre-announcing that they're not going to recommend Microsoft Surface this year, they're basically saying the product's reliability in previous years was so low that even if the new model turns out to be 100% reliable in their survey, its multi-year sliding window average will still be so low that they can't recommend the product.
You can see this in action in their auto reports. In deference to their subscriber base (who is typically clueless about statistics) they won't throw around terms like confidence interval and standard deviation. But some of the less-popular cars will have an asterisk saying they have an insufficient sample size. Even if they do have a big enough sample size, I actually prefer the sliding window method (with decreasing weighting the older the data is). It avoids the situation where with the new year, everyone's slate is wiped clean. If you have a history of making crappy products, it makes it harder for you to pull yourself out of the pit you've dug yourself into.
Personally, I really like the specs of the Surface of Surface Books. But I won't touch them for the simple reason that they're impossible to repair. If you're gonna buy one, make sure you get a multi-year extended warranty with it.