You only actually need one of them to be empty. The purpose of the empty group for the encryption domain is to break the domain-based logic for flagging "interesting" traffic. That logic says if the source is in my encryption domain and the destination is in a peer's encryption domain, flag the connection for encryption to that peer. This matching is done before any NAT can be performed. By setting one encryption domain to an empty group, either the source isn't in yours, or the destination isn't in the peer's.
That leaves the traffic free to route "out" the VTI, which is what triggers encryption with a route-based VPN.
The encryption domains are also used for a sort of anti spoofing. If you've ever seen "According to the policy, the packet should not have been decrypted", that means the packet was decrypted from a particular peer, but either the source was not in that peer's encryption domain or the destination was not in your encryption domain. Similarly, "Received a cleartext packet inside an encrypted connection" means the packet was received in the clear, but the source is in some other firewall's encryption domain and the destination is in yours, so it should have come over a VPN between those peers. Route-based VPNs' use of an empty group for one or both of the encryption domains means you should never see either of these, though you may still see normal antispoofing drops.
