You have an interesting definition of failure.
And C got the floating point operations B did not have, and was actually used for lots of numerical work for a while (and it still is, up to some point), but nobody was totally happy with it so we got C++, new Fortran versions, Python, now Julia... Although it always is an option due to the availability and quality of C compilers, I would not say that C succeeded in becoming the language of choice for scientific computing.
Of course, this does not mean that C or Fortran are failures.
So, "at least one operating system programmed in Fortran" doesn't counter "Fortran hasn't been really used for systems programming", if that operating system is some irrelevant research project, or (as in here) a commercial niche system, that maybe ever reached 0.00001% of the OS users space.
The first section of the linked article says that Fortran "was an odd choice from a pure computer science standpoint", that they "developed hardware optimized to run FORTRAN", and that large parts of the OS (in particular, the kernel) were eventually rewritten in other languages.
This is, in my opinion, a great example of how Fortran could not compete with C as a systems programming language.
> All arithmetic in B is integer, unless special functions are written. There is no equivalent of the Fortran IJKLMN convention, no floating point, no data types, no type conversions, and no type checking. Users of double-precision complex will have to fend for themselves. (But see ->32 for how to call Fortran subroutines.)
I wonder if it would have been beneficial to continue with this separation. Fortran and C started trying to do everything and they failed. Fortran was never a contender for C for systems programming, and C has not moved in the right direction to be better than Fortran for numerical computing.