(I seriously believe that any attempt to define FP will end in tears. It's a marketing term and has no real likelihood of producing legitimate technical value.)
FP as a programming style to which a programmer may adhere to a greater or less degree is easy to define. FP as a binary is-or-is-not characteristic of programming languages is, well, a misguided concept. To the extent it makes sense to talk about FP -- or OOP, or any other style -- as a feature of a programming language, it is more in terms of the degree to, or better the particular manner in, which it facilitates that style of programming, rather than a binary descriptor.
Functional languages are languages used predominantly by functional programmers.
Being functional probably has more to do with how a language is used than the set of features it boasts.
Can you do FP in any language? Maybe not COBOL, but you could do it in C with function pointers. But it's still easier to do it in Haskell than in C, and it's still easier to do imperative programming in C than in Haskell.
So you have a point. But languages have different characteristics, too, and those differences let us classify languages, if not as "FP" and "non-FP", at least as "more FP" and "less FP".