Your absolutely right and this is a statement that I do agree with. I think a difficulty in this area is that no one really defines what they mean by "software patent". In your example what is considered a software patent? the process of conducting the transaction? The software that does the transaction? the software that runs the whole process? I would guess the latter, and in that case your patent would likely not cover specific code but rather the abstract idea of whatever you need to do to go from point a to point b. That means I could use the software that conducts the transaction to do something else without violating your patent.
When looking at the applied math argument though, you have to consider, just how much application is necessary? This is exactly one of the questions the Supreme Court has looked at. In their ruling in Mayo (Which is viewed as essentially invalidating diagnostic medical patents as nothing more than patenting math) the court seemed to set up a sliding scale between the Flook case and the Diehr case: both cases involving applied math and both involving business method patents.
Diehr dealt with a patent on a process for curing rubber in which the patent involved using the Arrhenius equation to control the amount of time rubber was left to cure in a mold. The Supreme Court found that allowing the patent did not exclude others from using the equation for different purposes and processes.
Flook involved using a new equation to calculate alarm limits during a chemical reaction. The Court reasoned that this was merely patenting the formula and allowing the patent meant others could not use the same formula because the patent added nothing to the formula other than updating the inputs.
It's hard to reconcile these and the Court gives little guidance on how to apply their "inventive concept" test which is essentially what you described: The patent must USE a mathematical algorithm and not exclude other from the use of that algorithm in different contexts or for different purposes. With the recent rulings in Bilski and Prometheus, it seems the court is beginning to swing further in the direction of the need for some real inventive step or additional something beyond the software or math that underlies the process. Simply applying inputs, running an equation, and reading results is not going to be enough.