I agree with your points about code written in C by programmers, however...
You can't do much better than C by going to assembler, so C is a good target language. But much code written in many languaged could be improved by an automated translation into C with concurrent optimization. (The optimization needs to be done before it hits C, because you lose too much information in the process of translation.) The automatic translation avoids the difficulties of bounds checking, etc. If done properly it would (optionally) implement bounds checking wherever it couldn't prove that it could be omitted, etc.
Also, native C garbage collectors are inherently inefficient, because the C language doesn't reliably separate pointers from integers (etc.). But the translator, knowing the original language, could do this much more efficiently in the process of compiling from the original language to C.
So there's no reason not to standardize on C at the base level. And, of course, C compilers can optimize the C code.
FWIW, I dislike programming in C for multiple reasons. One of them is how it handles unicode. Another is the difficulty of implementing "class instance variables". (Class variables are easy, though. You just have one "class" per file, and static variables are equivalent of class variables.) I also prefer to have a good garbage collector. I dislike using pointers to reference structures. (In C++ I prefer to pass references as parameters rather than pointers to structs.) Etc. Of the languages that I'm familiar with, D is my preferred language, but it's missing a lot of library support, so I often use Python. Vala would be an excellent choice, if it coudl ever get it's documentation even to a beta level. (Do note that valac, the Vala compiler, has an option to allow you to generate C code.)