You are right the names for the fields and their description are misleading.

PublisherAddress is used to describe the destination address, where the NetworkMessage shall be transmitted to at the publisher (publish to this address).
SubscriberAddress is used to describe the receiving address, where the NetworkMessage shall be received at the subscriber (receive at that address).

For redundancy of the data … it lacks some description, why this element exists:
The communication model (i.e. for PubSub the CommunicationModelConfig) is a very complex structure, which needs deep knowledge about PubSub. However, to support what is stated in 6.8.1 (a Client looking at the ConnectionConfigurationSet and doing some modifications), the PubSubCommunicationFlowConfiguration serves exactly for that purpose – to give a Client the opportunity to tweak some selected parameters of the communication model (e.g., addresses, interval, security or QoS), without need to get into deep understanding of PubSub. In addition the PubSubCommunicationFlowConfiguration offers SelectionLists (prepared by the engineering tool generating the ConnectionConfigurationSet), where the Client can select from, whereas the PubSubConfiguration2DataType offers a single plain value.

Means, if the Client changes anything at the PubSubCommunicationFlowConfiguration, those changes are reflected at the appropriate element in the PubSubCommunicationModelConfiguration. For example changing the “PublisherAddress” will be used to set the Address in the WriterGroup (if unicast) or PubSubConnection (if multicast). Changing the PublishingInterval will be used to change the PublishingInterval in the WriterGroup and in the DataSetReaders …

Updated spec

Reviewed updates in call, agreed to changes and closed issue