Pip is correct about the remote default gateway. This is a specific choice made by the VPN administrators. It is sometimes for connectivity reasons -- ie connecting to non-local systems across a corporate network -- and it is sometimes sold as a "security" feature, in that systems connected to the VPN either cannot contact the internet or must do so through the corporate internet presence where presumably such activity can be logged.
I am not aware of any circumstances where fiddling with the local routing table gets any lasting effects -- you would probably have to write a script to to repeat the fiddling during each connection to the VPN.
Brent's solution is perhaps the easiest. I personally have a VM for each customer network just so I can be connected to multiple VPNs (mostly SSL-based, but some Cisco clients too) at the same time without having to worry about conflicts.