I ran across an issue whereas the data extract for a table was missing records and, in turn, the extracted total did not equal the total within the visualization. After a bit of analysis, it appears ...
I ran another test by taking the dataset model, taking a small subset of the data and switching it to a full import storage mode. The visualization then rendered all rows and the extracts total aligned with the visualizations reported total. However, a full import storage mode is not feasible due to the underlying volume and overall size of this dataset.