The main reason why you would want to do something like this is if you wanted to have traffic from loopback to loopback that didn't have an MPLS label. If the advertisement filter is set to only advertise the MPLS loopbacks and not the other loopback, you can use the other loopback for cases where you don't want MPLS labels to be applied to the packet.
For instance, suppose you wanted two routers with MPLS loopbacks to have an EoIP or GRE tunnel running from one to the other for some reason, instead of using VPLS. A use case for this would be if you need to act on DSCP for a tunnel since VPLS has no "inherit DSCP" option like EoIP does and so therefore the DSCP for packets entering the tunnel cannot easily be read. In this case, if you have two loopbacks and only one is set to advertise via the advertisement filter, you could use the non-MPLS loopback as the EoIP or GRE tunnel endpoint IP, and then your GRE packets would have no MPLS label. If you only have one loopback for both, your GRE packets would have MPLS labels applied and be label switched, preventing QoS rules matching DSCP from working on intermediate routers.