Because this specifically excludes retransmissions, it's a clean count of genuinely new authentication attempts sent to this server. It gives the admin an accurate login-attempt volume baseline to compare against the accepts, rejects, and timeouts counters for the same server when calculating a real failure rate. For example, 500 access-requests against 495 accepts and 5 rejects on this server describes a healthy roughly 99% success rate, cleanly separate from whatever retransmission noise occurred underneath.