Unlikely to be opposing trends

The real numbers are probably commercial secrets. Particularly if they're significantly different to the numbers vendors in other sectors are used to. They're not the "secret sauce" but they are valuable information.

It would be interesting to see what figures looked like for Zeta, a BeOS derivative (which subsequently turned out to be unlicensed, but that's not the story we're interested in here) which was sold into the general consumer market via TV shopping channels. Initially high sales figures were quoted for Zeta but that didn't translate into money to keep the company behind it alive. Obviously there could be lots of reasons for that, but a high return rate would certainly help explain it. I've heard rumours that many customers felt it hadn't been clear enough what they were getting. Unlike these "netbooks" the Zeta OS was mostly sold as a pure software package, so presumably no-one was expecting Windows.

Of course it could just be that other characteristics are different in either the Linux-wanting community (e.g. Linux users expect hardware to feel well-built, find the netbook cheap-feeling and return it, or Linux users are more likely to take advantage of "no questions asked" return policies to try out products from a bricks and mortar store) or among people who emphasise price and might thus buy a Linux netbook because it was $10 cheaper than the Windows version. But in a commercial sense this doesn't matter - this is an unexpected cost and if the number crunchers decide that it's too high, no more Linux on netbooks.

Coca-Cola knew from blind testing that New Coke was better than Classic. But once it became apparent that the same customers who chose New Coke in taste tests were demanding Classic from retailers that's what Coca-Cola had to make. In this sense it's true what they say, "the customer is always right".