You're reading something into the title that isn't there. The author says "7 bits aren't enough for 2-digit accuracy" and you're reading it as "7 bits aren't enough to represent 2-digit numbers." The post is about converting base 10 floating point numbers (with 2 digits accuracy) to base 2 floating point numbers and back, not representation of integers.
I don't think he's reading too much into the title. He equally well showed that it's possible to encode 2 base-10 digits of accuracy in base-2 fixed point numbers and back. Nowhere does the title mention floating point numbers, which is just one arbitrary (but popular) format among a number of possibilities to encode numbers with chosen accuracy.
fair enough; maybe it's too far outside my area of expertise. OTOH I never expect round-trip conversions to be perfect and couldn't care less whether there are small errors, and would use a different representation if I did.
