Arve Bersvendsen has written an interesting post in his blog about misleading browser stats. As a quick summary, Opera's aggressive caching could lead to fewer hits compared to other browsers with less aggressive caching (IE and Firefox). This is even worse for text-only browsers like Lynx, as they typically don't request external style sheets and images at all.
Read more in Arve's post about misleading browser statistics.
In addition, Opera identifies as MSIE by default, and even though "Opera" is part of the useragent string, many statistics sites fail to detect Opera. In Opera 8, it is even possible to completely cloak Opera, and that means that statistics for those sites show no Opera users at all.
An Archive From Behind the Trenches 
Surely Opera has been in contact with the peddlers of these stats – they seem to be large firms(?), and there can't be that many of them. I agree with most of the assessments Arve made, but this discussion has been going on for a while now, surely you've contacted these people to set them straight?
Because it makes more sense to me that you'd want to complain about it rather than help them get it right and sign off on their stats – if they still ended up saying your (our) market share was as tiny as it is now. If you're willing to contact sites that write broken code, wouldn't you want to set these stats people straight? Or is it more difficult than I could possibly imagine?
And as I mentioned on Arve's site, are you sure they count by hits rather than unique users? Since that would invalidate some of the arguments.
It is undercounted when you look at 'hits', but if you cound unique visitors, one hit is enough (other consecutive hits are ignored) and Opera's caching shouldn't skew those stats.