Cutting this short, simply because it's late and I finished the application I was drafting, but I was enjoying this debate...
You're a patent lawyer, where would your boundary be?
The original prohibitions on patent eligibility expressed by the Supreme Court of "laws of nature, natural phenomenon, and abstract ideas" were not in the statute, but were a clumsy interpretation as SCOTUS wrestled with a more fundamental problem: patents allow the patent owner to seek an injunction, ordering people to stop infringing... If the patent claims a previously undiscovered law of nature such as, say, gravity, how do you order people not to fall? Similarly, if the patent claims a mere abstract idea or mathematical algorithm that could be done in someone's head, then how do your order people not to think?
For example, if the patent claims a diagnostic method involving noticing that a patient has a blood level of X of n ppm, and determining that therefore they have disease Y, and I tell you I have a patient with a blood level of X of n+1 ppm, you just infringed the patent, simply by recognizing that fact. How can a court order you not to think of something? Or, conversely, how can a court fairly order you to pay royalties for thinking, or for falling, or otherwise involuntarily infringing a patent?
... which gets to the real point. Patents shouldn't be able to claim something that can be infringed involuntarily - whether that's a law of nature that one uses constantly without realizing it, or a mental algorithm that one can infringe merely by being told about it. So, that's my boundary - if the claim can be infringed through mental steps alone or otherwise involuntarily, then it should not be patent eligible. If the claim requires programming a computer to do something, however, then one may avoid infringement simply by not programming the computer to do that thing.
Mind you, this is all about patent eligibility under 35 USC 101... We've been talking about novelty under 102 and nonobviousness under 103, and those are separate and independent requirements. I agree that the quality of examination needs to be better. Part of why courts keep falling back to 101 for invaliding patents is because they can't believe that the patent isn't obvious under 103, so they refuse to hold them valid, but they don't have any good prior art to point to, so they instead wave their hands and claim it's abstract, without ever defining what "abstract" means. That's just bad law.