What I believe the difference is that someone could send a SegWit transaction to a non-SegWit node that was never valid, but the node would treat said transaction as if it were valid.
That is possible, but the node still would not accept the transaction because segwit output are considered non standard. Such a transaction would thus be considered non standard and rejected by the node until it is included into a block.
If someone who has not upgraded to SegWit receives a SegWit transaction, are they able to spend outputs from said transaction?
There are in fact no such things as segwit transactions. There are transactions which spend from segwit output types. Because those output types are created by the receiver of a transaction, then it is impossible for a non segwit wallet to receive a segwit output to that wallet because there are no segwit outputs associated with the wallet. A transaction can spend from a segwit output types and spend to a legacy output type and that is perfectly fine and will be considered valid by the receiving non-segwit wallet (of course it has to be confirmed first because said transaction would be non-standard).
Is it true that if Segwit gets activated that Bitcoin addresses will start with "3" than the normal "1"? How are we going to be forced to make the switch with the most hassle free way possible?
There already addresses that start with a '3' and they have been standard for many years already. These addresses are called Pay to Script Hash (p2sh) addresses and are most commonly used for multisig addresses.