Kevin,
The CAD-BE is Java based and as such cannot put the NIC in promiscuous mode to listed to the RTP stream so there isn't any way for it to forward traffic to the supervisor.
As you mentioned, using SPAN to capture traffic on the Voice VLAN is the other method. Since you have L3 network you have two options the way I see it:
Option 1. Use RSPAN to forward the RTP traffic across the L3 network to your server.
Option 2. Put a server local to the agents you wish to monitor and record.
Please help us make the communities better. Rate helpful posts!