From that it looks like a variety of models have all reached conclusions of 1-2%.

I can't think of a strong reason to suggest that it would be higher. The argument is made on that page is that it is strong in places like China, though my understanding is that is not correct. Plenty of PCs in china may sell with Linux (or freeDOS) but my understanding is that they are always ubiquitously flattened and replaced with pirate Windows.

The point was, that those models aren't comprehensive. So using that number is not better than using higher estimation. In reality - exact numbers are hard to come up with, but you can conclude that 1% is underestimation using indirect methods.

What indirect methods show that as an underestimation?

Evaluating the skew in the web counters. Again - it's not a clear proof. Rather it's a reason not to take web counters like a proof.

Ok, but then there's no reason to suspect it would necessarily be higher rather than lower.