I agree with #3 and #4, but my background is in Sanskrit and Hindi (and scripts for related languages like Punjabi or Bengali). There were set rules for transliteration so that each phoneme in Devanagari (and related scripts) was represented by a unique Roman character, though accent marks were sometimes used for some sounds.
Translation, though, was a process of interpreting the syntax and semantics of a language so that it made sense in the language you are rewriting it in. For example, Hindi uses dative constructions. To express that you're sick, you'd say something like, "there is sickness to me." But this makes no sense in English, so when you translate you have to come up with a sentence that means something like the intention in the Hindi phrasing - for example, "I feel sick."
Transliteration to me seemed kind of like a straightforward change in the way the language is encoded. You could do it if you had your secret decoder ring. But translation is more of an art than a science, because there are not always (or even often) one-to-one analogs from one language to another, and it takes some creativity to find a phrase in one language that means what the other language is saying.