Tuning Myths

Matt's post (No 9) in the Query Execution Plan discussion sparked a thought. 

What are the myths about tuning? Those things that somehow seem to have become tuning "golden bullets".

I'll start off with

1) Avoid having Full table Scans at all costs.

2) In the same vein but the other way round -  Always get the plan to use indexes.

3) You can tune a system by looking at the hit ratios.

4) You can meaningfully compare the cost of different plans.

5) If you give a hint it is always followed.

My thought was to initially just list the myths, rather than go into details, as it will provoke more thought