With DCF, apparently if Station B sends an RTS to station A, then C delays for NAV(RTS) but cancels if it doesn't detect a CTS. If this were the case it could then sense for DIFS but detect data from B. I'm assuming that it would ignore the fact the medium is busy and continue to transmit an RTS to D because its worked out the data from B will be received by A regardless of interference. I assume B ignores any activity on its receiver when it is transmitting. Similar to how it ignores activity when it is deferring access.
Imagine node F is introduced above and between B and C such that B, F and C are all in each other's range. If B sends an RTS to F and C picks it up and delays for NAV (RTS), but there is a corruption at F so it doesn't send a CTS, but rather it waits for EIFS; then wouldn't C assume F is a hidden node, sense for DIFS and then send its RTS to the node it wishes to communicate with (unless it is also B, because it would assume it busy)? This is potentially an example of the exposed node algorithm manipulating a scenario without an exposed node.
To make things stranger, on the MACAW Wikipedia page:
An overhearing station (node F), which might have received RTS and DS but not CTS, defers its transmissions until after the ACK frame should have been received plus a random time
It seems that MACAW doesn't support cancelling NAV(RTS) if it doesn't receive a CTS and instead waits the whole NAV. So MACAW doesn't mitigate the exposed node problem?