In the question's example, C is acting a Man-in-the-Middle (MitM), impersonating B with respect to A (at the communication level) when receiving {RB,B}K, and impersonating A with respect to B (similarly) when sending {RB,B}K. Absent measures not described, B will accept the message {RB,B}K sent by C as if it came directly from A.
Such possibility of a MitM is a standard hypothesis in cryptography and protocol design, because it's it's physically possible in the overwhelming majority of communication means in use today: wires, optical fiber, radio (including satellite).
Whether "B will authenticate C instead of A" is a matter of convention and context, and these vary.
Absent other attack¹, the protocol insures that A participated in the exchange, and I would not say that the authentication property of the protocol limited to what's described in the question² is broken or useless. For example if A is a remote control for a device B, then the remote was involved within the time period between generating RB and accepting {RB,B}K. If further B verifies that this time period is as most $t$, then B has the insurance A is within distance $t\,c/2$, where $c$ is the speed of light.
¹ Attacks are possible or not depending on properties of the encryption used, and on how parties with knowledge of key K (including A and B) use it beyond what's described in the question.
² Things would be quite different if the question ended in: having authenticated A, B proceeds to communicate.