I think that Jeff describes it well. A bit more detail is provided in https://hydroacoustics.usgs.gov/memos/OSW2009-02.pdf. I have pasted in the pertinent wording below.
For each ensemble, WinRiver calculates a depth below which velocities measured by the ADCP are not used due to possible errors caused by side-lobe interference. This depth is referred to as the side-lobe cutoff. In previous versions of WinRiver (2.02 or earlier) the sidelobe cutoff was calculated as 6% of the depth, computed as the mean of all valid beam depths for an ensemble (figure 1a). The new side-lobe cutoff is calculated as 6% of the shallowest beam depth in an ensemble (figure 1b). For example, the side lobe cutoff computed for ensemble 78 in figure 1 for prior versions of WinRiver II is 2.4 ft. The beam depths measured for this ensemble were 1.9, 3.5, 3.3, and 1.7 ft. The new side lobe cutoff, computed using WinRiver II version 2.03 and 2.04 is 1.6 ft. The change in the side-lobe cutoff may result in fewer valid depth cells in ensembles near sloping banks and in cross sections with irregular or rough streambeds. As a consequence, the middle (measured portion of the cross-section) discharge will decrease and the bottom (unmeasured or extrapolated portion of the cross-section) discharge will increase.
OSW has used the new version of WinRiver II to reprocess discharge measurements from Oberg and Mueller (2007) as well as other available measurements. Based on the discharge measurements reviewed to date, the new side-lobe cutoff has resulted in a median difference in discharge of +0.5 percent. The typical change in discharge was less than 1 percent. However, for about 4 percent of the discharge measurements reviewed the discharge changed by more than 5 percent. The largest changes were found for measurements made in shallow uneven cross sections that would typically be considered poor measurement sections. These cross sections are characterized by mean differences of beam depths for individual ensembles ranging from 17 to 32 percent.
Measurements collected in the 2009 Water Year with a mean difference of beam depths in individual ensembles exceeding 15% or a mean number of valid depth cells per ensemble less than 4 must be reprocessed using version 2.04.