They are reporting what the raw gender wage gap is for weekly pay of full time workers (>35 hours). It doesn't take into account the actual number of hours worked, the amount of overtime worked, the risk of physical injury of the job or any other conditions of employment, union representation, gaps in employment, experience, tenure, the industry involved, or the preference for fringe benefits in lieu of pay.
Whatever portion of the gender pay gap that's due to illegal discrimination is almost impossible to measure, so economists factor out that which is relatively easy to measure. The Consad meta-analysis is the most comprehensive gender wage gap study we have, and puts the actual unexplained portion of the pay gap at between 4.8 and 7.1 percent. Within that 4.8 to 7.1% you have a number of other factors which are difficult to measure, one of which is illegal gender discrimination.
Additional portions of the raw gender wage gap are attributable to other explanatory factors that have been identified in the existing economic literature, but cannot be analyzed satisfactorily using only data from the 2007 CPS. Those factors include, for example, health insurance, other fringe benefits, and detailed features of overtime work, which are sources of wage adjustments that compensate specific groups of workers for benefits or duties that disproportionately affect them. Analysis of such compensating wage adjustments generally requires data from several independent and, often, specialized sources.
So when the question is 'how much of the gender wage gap is due to illegal discrimination?' The answer is, very little.