Discussing the CAP theorem is a discussion on what is theoretically possible.
All I am claiming is that it is theoretically possible to make a CP system that, in practice, is highly available. That does not mean one actually exists.
I think the better argument against the CAP theorem is that while it is true, is it not really that useful. If you have a database with nodes in three data centers, it is much more likely that one of the data centers becomes disconnected (partitioned) from the rest of the internet, rather than that the internet itself is partitioned. In this case, a CP system can still remain available to the internet, it just can't also remain available to the nodes within the disconnected data center.
The current RavenDB documentation still warns against using "system transactions" because of performance reasons (http://ravendb.net/docs/client-api/advanced/transaction-supp...), so I think it is still fair to say that RavenDB was not designed for applications requiring high performance cross node transactions.
Presumably you will want to retry conflicting transactions, so you generally would not count them towards your throughput.
For example if I commit 100 transactions per second, and 90% of them return conflicts, I am only successfully committing 10 transactions per second.