I read there re cases where the final exponentation on elliptic curves pairings is easy to invert, but is it true?

I read that for some curve this is possible with the text being specifically, if `$\gcd((p^k-1)/r, r) = 1$, the final exponentiation is a bijection on the r-torsion and can be inverted by computing the modular inverse of the exponent modulo r`.

But is it true as it seems such assertion will always be true to me for prime order, and if yes what does it means?