Yep, in fact, I think that is really what killed Ada use.   Because of the need for embedded support and most of the small processors did not have good Ada support, but did have C and assembler, a lot of USG contracts applied and got variances.    But the start of the 90s, it was pretty clear, the idea behind Ada and a standard language for the USG was a lesson of theory vs. practical reality.

Ada had a huge spike on the Mainframes and Minis because when it envisioned (in the mid-70s) that was the target processor.

I used to be friends with the then chief SW guy at Raytheon who lead the Patriot missile SW development during those years (we lost him a few year ago due to massive heart attack).  But he made me understand why Ada was created.  At the time, Raytheon was doing support for the Polaris submarine missiles.   They did not have the full source.  It was all patches.  DoD wanted a programming language that they could use for both specification and deployment.   They wanted the specs to be able to last.   And an interesting idea.

But as you point out, as time went one, more and more of the code went from being in large systems into embedded micros and they were back to the same problem.  The lacked tools to take Ada to deployment.  So they specs might have been written in Ada 'pseudo-code', it was all done in C and Assembler.

Clem