That's certainly true, although notation like `fib_n = O(phi^n)` (rather than `fib_n \in O(phi^n)`) is so ingrained that it's probably too late to fight it. (I seem to remember that Knuth says something to this effect.)

In that spirit, one can adopt a sort of compromise notation: `O(phi^n) = o(2^n)` (where `=` should really be `\subseteq`).