An example:
We have 3 web servers for a specific server, if one server is down, the service is affected, but not down so the SLA is still met then the % should not decrease. If the customer wants to declare the service down if 2 of 3 webservers are down, then the clock should run when those servers are down.

Another example:
The service EMAIL is composed by 3 servers, an POP3, an SMPT server and a SMTP/LDAP server. If a disk of the POP3 server is getting full, the service will not be green, since there is a trigger for the server, but the other servers are OK. If one of the SMTP services running on the machines is down, the other can act as contingency, so the service is not down also. But, if the LDAP is down, the service is down, because users cannot authenticate to send/receive mails, so in this case the SLA should be affected.