In your scenario most risks are associated with certificate management, because self-signed certificate offers zero manageability. For instance, certificate replacement, update, revocation. The lack of manageability leads to increased security risks.
For example, you use self-signed certificates for b2b communications. Internet-connected endpoints are more vulnerable to different attacks and there is quite big chance that the certificate/key is leaked/compromised. This can be a vulnerability in web app software, web server, OS, etc..
In order to replace the certificate, you will have to inform your partners about the incident, ask them to remove old certificate and install new self-signed certificate to their trust store. This process is very slow and during this timespan your partners will trust your compromised certificate, because of this delay.
When deploying standard PKI solution, it is easier to provide root certificate's adequate security level, because attack surface is very tiny. CA is not connected directly to internet and has very limited access within private network. In this case, root CA compromise is very unlikely. And if your web server got compromised (you can't afford the same security level like in CA server) you just revoke the certificate and replace the certificate on web server on a fly. And partners won't experience security or outage issues.