Low probability of enforcement is evidence that the system is not working correctly, for a couple of reasons:
You can't simply factor out the probability of enforcement, because uncertainty also has a cost. And it's a cost to the system as a whole, not just a transfer between parties, because uncertainty causes friction. (It may seem peculiar to say that uncertainty is a problem for entrepreneurs - aren't they supposed to be all about taking risks? But we want them to spend their risk-taking capacity on innovation, not working round legal uncertainty).
This degree of uncertainty over property ownership is not seen in other areas of property. Yes, discounting by risk occurs in real property, but the vast majority of real properties - at least in the developed west - have much more certain ownership; with discounting only used to get rid of residual risk. We have seen from the recent mortgage debacle that it causes huge stress if too much uncertainty is allowed into the system. In developing countries, uncertainty over property ownership is a huge drag on the economy. Even in patents, in other industries, such as pharmaceuticals, small companies rely much less on assuming that enforcement will not occur, and actually do licence. That is why the issue in pharma is whether there is a 'patent anticommons'.
You make the same argument that Paul Graham makes (http://www.paulgraham.com/softwarepatents.html), that trolls are content to wait for the bigger fish to buy up the smaller ones. We can assume that he knows what he is talking about. But we can't assume that it will continue that way. Not all smaller companies are bought by bigger ones; in fact, small companies make up a large proportion of the economy. Trolls will want to monetise them if they can, and we can expect them to become more efficient at doing so over time (eg, via exchanges such as http://www.ipxi.com, which is implementing the ideas in Lemley and Myhrvold's "How to make a Patent Market (http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1012726)). Also, it is not necessarily the case that infringing a patent simply incurs some percentage 'tax' on profits. In a perfectly efficent world, each troll would negotiate separately with each infringer, but that can't happen, so we see fixed prices which prevent certain uses, such as in the MPEG pool. So the liability created by patents can make some businesses uneconomic, and this will become a larger effect once smaller companies are drawn into the system.
Secondly, the uncertainty is not just due to the low probability of enforcement, but due to the inaccuracy of determining whether you are infringing a patent. This inaccuracy dilutes the incentive effect that is supposed to be the reason for patents.
You are right that we can't obtain perfect accuracy, but the interesting question is how much does it const to obtain enough certainty for the system to actually work. That's why I suggest measuring how much information you can get for a given cost.
Also, from a political point of view, it's worth challenging the simple narrative that proponents of software patents push - in which software patents are a 'no-brainer' benefit to the economy. Exposing the real situation is well worth doing.