An important difference is that some UARTs (16550 notably) will stop receiving characters immediately if their host instructs them to set DSR to be inactive. In contrast, characters will still be received if CTS is inactive. I believe that the intention here is that DSR indicates that the device is no longer listening and so sending any further characters is pointless, while CTS indicates that a buffer is getting full; the latter allows for a certain amount of 'skid' where the flow control line changed state between the DTE sampling it and the next character being transmitted. In (relatively) later devices that support a hardware FIFO it's possible that a number of characters could be transmitted after the DCE has set CTS to be inactive.