Well, of course we need to specify the destination address. In the MPLS case, we would signal the router serving us that we wish to talk to a certain address, and the router would send back a label ID that corresponds to that connection. (While the destination addresses are global, the label IDs can be reused per pair of devices, but that is besides the point). At this point, the path is set up and cannot really be "messed with" and you reference it by the label ID.
The security benefit is that the routing mechanism is invisible to the end user. He needs to specify the destination and the rest of the connection is up to the network.
Of course, the other benefits are efficiency and traffic engineering. With the network being aware of the actual connections (unlike with TCP, where packets are essentially disjoint from a router's point of view), it is relatively easy to provide features like bandwidth reservation, QoS guarantees, etc. And the actual switching process for circuit switching is a lot more efficient. It is far easier for a router to perform a label lookup and then push/pop/swap labels than it is to carry out the longest prefix match lookup. In fact, such technology is already used internally by some ISPs, but it is not available globally or end-to-end.

How is that vague and meaningless? One of the issues is that users are able to exercise direct control over the network through the same "port" which is used to send data. Sure, you would have to send some messages to set up a connection to your destination, but that control action should be done by the network, as it sees fit, instead of letting the user set the source and destination addresses on every packet. The user should be able to ask the network to set up a connection and then send data, and the network handles all the internal operations. Going back to the phone analogy, are you able to pick up the phone and make a call to someone while faking your own phone number? No. Are you able to place a call to and directly manipulate telephone switching equipment? Not since decades ago. Are you able to hijack someone's conversation, or force a specific path for your phone call? I don't think so.
The original design of the internet did not anticipate the need for isolated control, management, and data planes. There was just no reason to do it back then. But with 30 years of development and growth, things have changed...

There has been some talk about separating the control plane from the data plane (ie, packet header from data). The phone network had its share of unsecurities when they were using in-band signalling, but since the two planes were separated, phones became far more secure. The same technique can be applied to the data network. If we separate the control information from the actual user data, we may achieve better security, as it would thwart any attempts to mess with the packet header, redirection attacks, prefix hijacks, or any of that other garbage. And the technology already exists. Look at MPLS- your computer can signal the upstream equipment to set up a connection to a specific address, and all you have to do is send forward data with the given label ID. The technology is already being used to route traffic within ISPs, but the security benefits of it won't really materialize until it's pushed out to the user level. Of course, good luck getting everyone switched over.

Thank you sir. It is the first thing I hit ^F for when this page loaded.

That doesn't quite constitute using (overall) signal strength, and neither is it the primary location method. Yes, it is possible to use the carrier phase information as well as the L2 carrier phase (and L1/L2 discrepancy) to get a more accurate fix, but this information is only used to adjust the TDOAs of the PRN signals and compensate for varying ionospheric delays. Signal strength of each satellite is much more affected by random low clouds and even the receiver's immediate environment, than by distance from the satellite. In fact, if your receiver provides an SNR readout for each satellite, you can get an idea about just how dramatically these values are affected by, say, a tree that partially obscures a portion of the sky.