|Think about Loose Coupling|
You are right, of course.
Problem is, in cryptography there are probably only a handful of people worldwide for each algorithm that can actually tell you which parts of the system you can leave out and while only degrading the encryption product from "secure" to "still reasonable secure". Even a single, small error can lead to "not secure at all".
So, what i meant is, a non-specialist should assume that he/she can only turn the "secure" switch on and off but has probably not the knowledge to tweak it into a point somewhere inbetween.
As Dan Kaminsky said about the Debian SSL desaster (the non-randomness introduced to fix a memory access bug that wasn't one): "You know, it's not a private key if there's only 1 out of 65.000 of them". Youtube Video at 21:00. There is more on Dan's Blog.
"I know what i'm doing! Look, what could possibly go wrong? All i have to pull this lever like so, and then press this button here like ArghhhhhaaAaAAAaaagraaaAAaa!!!"
In reply to Re^8: Your random numbers are not that random (UtS,L)