You only need the ports associated with the VPN forwarded on to the server. Read back over both posts, but still not sure if you are dealing with PPTP or IPSec. For PPTP, you should have TCP/1723 forwarded and PPTP pass-through enabled (called GRE protocol in a few routers, don't think you have anything that applies to). If you are using IPSec, it would be UDP/500 with IPSec pass-through enabled.
You don't need anything else open to access resources on the remote network, as anything you do over the VPN is wrapped up in a packet that passes through the router/firewall (either in a PPTP or IPSec packet). The firewall isn't aware/doesn't care what you are sending in the VPN packets. If the VPN is connecting, and it sounds like it is, you have enough ports open, at least at any point between the two computers participating in the VPN.
You should allow ICMP requests to allow for troubleshooting. You may need to ping or traceroute to the router at some point. I generally setup rules to block ICMP after 50 requests in 30 seconds or 700 in 10 minutes -- allows for a single source to ping constantly, but kills a DOS attack at that level. Not a big issue.
It sounds like the VPN is running fine, your addresses are ok. Possible problems left at this point . . .
A) Firewall software on either the VPN server or VPN client. Again, this will vary depending upon what you have installed, but the VPN will create a virtual network interface, and anything coming in on this interface is VPN traffic and should be allowed. If you have software installed, and it allows you to further limit traffic by IP address, you may want to limit it to traffic coming in on the VPN interface that has a source address on the same subnet as the VPN. Would try without it first, just to get things running.
On that note, NT 4.0 does have IP filtering built in, but it uses a system wide policy, if you are able to ping the server from within the server side LAN, that is not an issue.
B) You could still have routing issues. I would discount this possibility, as you should see a 'no route to host' response to your ping or tracert after changing the IP addresses. This may become an issue if you try to work with other machines on the server side network, if you need to connect to a computer at 204.182.234.xxx other than the VPN server (which you are connecting to at 192.170.160.xxx, right?) routing will nedd to be configured to send the traffic over the VPN instead of the public internet. Note that if you connect to the VPN server with pcAnywhere and use other resources from there, this is not an issue as you are technically not connecting to those computers over the VPN, you are connected to the VPN server which is in turn connecting to those computers.
On that note, it would be interesting to see what happens if you were to try these from the VPN server to the VPN client if problems persist.
C) This is where I think the problem lies (at least with pcAnywhere). I don't work with pcAnywhere, but I got curious and had to poke around a bit. As I suspected, pcAnywhere only connects to the first configured IP address by default. In your case, this would be the server side LAN IP, which you will not be able to hit without the routing. It is going to be more efficient to connect to the VPN IP anyway, so you need to enable it there. Take a look here:
I would set it to allow connections on all network adapters, at least until you get it working to your satisfaction, then change it to listen only on the VPN IP if you feel the need later.
A lot of ifs/ands/buts here, but (there went another) I hope to have given you a good direction. Sorry if I repeat and/or forget details here and there, the threads run together sometimes.
Do post back if you need further clarification, or if something just doesn't work. You were close, and you still are. Has to be something minor, so hang in.