Mixed class widths
For some data sets, wider classes give a smoother histogram in some ranges of values (e.g. in the tail of a distribution) and narrower classes are better in other parts of the distribution (usually where there is greater density of values).
In a correctly drawn histogram, each value contributes the same area.
Histograms can be drawn with mixed class widths, but it would be badly misleading to make the rectangle heights equal to either the class frequency or relative frequency.