The article is specifically about round-tripping between base-10 and base-2 floating point because the precision of floating point is a very real issue if you're doing conversions. This is e.g. one of the reasons why decimal fixed point is preferable for many types of operations.
Fair enough that the title leaves it ambiguous, but the ambiguity is resolved in the first sentence of the post, and further clarified in a lot of detail over the next paragraph and a half.
The whole setup is a problem that shouldn't exist, because you're using 7-bit floats for an inappropriate range. When are you going to run into 7-bit floats as the default? If you're not thinking about it, I suppose this could happen if you scale it up, but it's less surprising to say "4862057400958023983457609283094859 can't be represented exactly in floats".
Fair enough that the title leaves it ambiguous, but the ambiguity is resolved in the first sentence of the post, and further clarified in a lot of detail over the next paragraph and a half.