Sure, just assume that 90% of the lawsuits brought by patent trolls lose at final judgment. You can't work backwards from that factoid and conclude that 90% of software patents are crap.
After several levels of filtering, only then is a case decided. And at each level, if the software patent i strong, the process ends.
Demand letter gets reviewed. Strong patent? Settle.
Initial lawsuit gets filed. Semi-strong patent? Cost-risk analyze defending the lawsuit and license if the numbers don't work.
Lawsuit continues. Weak but possible bad judgment? Settle.
So only the patents that the defendant looks at and decides that the risk of losing the suit, cost of the suit, _and_ the strength of the asserted patent are such that it makes sense to risk a final judgment are these 90% losses representative. Hardly a basis for a universal declaration that all software patents are weak and unenforceable.
Sure, folks may have a beef with the concept of software patents, but that is a separate issue.