The idea is that the shorter it takes to get into the bloodstream/brain, the peak effects are stronger, and the downregulation of neuron receptors is more acute and lasts longer (therefore requiring redoses to activate them).
Not sure about primary sources, but the professor talks about the routes of administration in the Drugs and Behavior class that's available on iTunes. In terms of addiction rates, this is something that Carl Hart researches and has been talking a lot about recently. He has written a bunch of articles and given some TED talks on the topic.
In terms of quoted figures for the addictiveness of various substances, you need to look at whether the studies you're looking at correct for the characteristics of the users. E.g. people who use heroin are much more likely to become addicted to drugs, which makes heroin look more addictive than it actually is. There certainly is some variance between drugs, but it's relatively small... like 9% vs 13%, but not like 2% vs 90%.
It actually doesn't, which is one of the key findings of addiction research. The reason is that people don't abuse drugs because of the physical discomfort of withdrawal, but rather because they have a psychological dependence on drug use. And that psychological dependence is caused by things like previous life trauma; any given drug is just an interchangeable object that's being fit into a slot.
The one drug-related variable that is important though is route of administration. Because addiction involves linking an action with a reward, the faster a drug takes effect the more likely a user will form a problem with that substance. So basically in terms of addictiveness:
smoking > mainlining > snorting > eating > topical
This is why different drugs seem like they have different addictiveness potentials.