It's the modeling of physical systems that its patented, not the physical object itself. Those patents usually protect the inventor from people creating any number of variations on their exact design because it's the approach to solving a particular problem that is being patented.
I think there are a lot of people who take an absolute position against software patents and end up going through all sorts of contortions to differentiate between why ideas in the physical realm should be protected and ideas in the software realm should not.
That said, I certainly agree that most software patents are far too obvious and not deserving of any sort of protection. The challenge, of course, is how to fix the system to afford protections to real innovators and not to those who simply connect the dots.
Algorithms are mathematical equations. They cannot be patented.
I know the mantra. I just don't buy it. Math provides the building blocks used to produce creative (and sometimes very innovative) solutions to problems in the form of software. Material properties and laws of physics similarly provide the building blocks used to produce creative solutions to problems in the form of physical devices or mechanisms.
Just because the basic building blocks used to formulate an innovation aren't patentable, it doesn't necessarily follow that the innovation itself shouldn't be patentable if it's truly innovative and worth protecting (which most software patents certainly aren't, but some undoubtedly are).
That leaves the patentability of math. Man I wish I had a time machine so I could go back in time and convince Pythagorous to patent his little triangle formula in the context of architecture. I could make a killing.
On the other hand, the numerous proofs that software is math are fairly trivial for anyone with a modest background in mathematics.