Really interesting topics i’ve been through recently regarding an internet infrastructure/vpn. We had issue during operation in which our PE didn’t import the prefix with let say local pref defined. This operation was made to improve flow repartition (outbound) from multiple datacenter.
As we didn’t understand what was happening a first place, we decided to rollback. Then I reproduced the whole infrastructure and figured out that there was (don’t know exactly why) but multiple PE with same RD for this specific VPN (internet). When I wanted to manipulate metrics from a specific route, my metric was not applying and I noticed the default one(100 for localpref).
I then labbed multiple case with basic mpls backbone and when trying specific manipulation, having different RD on each PE is recommended I guess.
I was not able to produce debug logs attached for explaining exactly why but I’d say as vpnv4 update, the device is already considering the route by his own and then my import wap was never matching at all.

Cisco documentation is not so clear about import map/export map in my opinion