In fact, the crucial point about the S flag is that it indicates that the source
IP address is usable and was the signalling source address. Revised text in
5.2.1 has this emphasis:

   Source addressing mode: A flag to indicate that the IP source address
      of the message is the same as the signaling source address, in
      which case replies to this message can be sent safely to this
      address.  It is not set if it was derived from the message routing
      information in the payload and this is different from the
      signalling source address.

(also changed in C.2.1).

The S flag indicates whether the source address in the IP header is derived from
the MRI or is the signalling source address. In the case where the originator is
an end system, these addresses are the same anyway (at least for a single-homed
node), so it is not clear which S value should be used.

In practice, the S flag is used for two purposes:
- to decide if errors can be sent (safely) to the IP source address
- to decide how the IP source address can be examined to detect the presence of
a GIMPS-unaware NAT

In that sense, it is almost certainly more helpful to know that the IP source
address is the signalling source address (in other words, for S to be set),
since if this is not known, there is nothing useful that the message receiver
can do. In other words, S SHOULD be set in messages from end systems.