Typically, estimates are generated for the purpose of deciding whether a project should start or continue.
I think it is vital, therefore, that each estimate detail line includes a percentage confidence factor.
I always break up projects (even little ones) into phases, tackling the highest (perceived) risk in the first phase. Confidence estimates usually get weaker for later phases which depend on the quality of results from the first phase(s).
Using this method I can present a realistic estimate with best and worst case values rather than an optimistic guess.