>>>But if you remove duplicates, it's less random?
>
>Maybe "prevent" is a better word? I think it's still random if you prevent duplicates by removing each selection, and making a random selection from the rest.

Less random means still random, but of poorer quality. If you do encryption or the like, it could be used as a vector to crack it. If we use prevented or removed ends up the same. Anyway, as you point out, the generation is bad random because it favours lower numbers due to the way it's generated.
:)