First, make sure you are not holding down the Shift key when you release the mouse button as you drop the attribute into a graph or summary table. The Shift key forces an attribute to be treated as categorical.
The most probable explanation is that at least one value for the attribute is not actually a number but a non-numeric string. Usually this is a data entry problem; e.g. someone typed the letter 'O' instead of the number '0,' or there are missing values represented by asterisks or some other non-numeric character.
One solution is to hold down the (Windows) or (Mac) key as you drop the attribute on the graph axis (or summary table). Using this key forces an attribute to be treated numerically. The non-numeric values are simply not displayed (or used in computations).
Another solution is to find the non-numeric values and correct them (or delete them). One way to do this is to make a case table of the collection and then to add a filter to it (use the Add Filter command in the Data menu). The formula for the filter should be
where xxx is the name of the attribute and '' means 'not'. You get it by pressing the not button in the formula editor keypad. After applying this filter, you should see only the cases that have a problem. Edit the values for these cases and remove the filter when done.