One of the last sections on "Problems of Honesty and Good Judgment" resonated with my experience in the working world. Oftentimes, misleading or difficult to understand graphs are included for the exact purpose of misleading or misdirecting an analyst away from topics that companies do not want people like myself (a credit analyst) to focus on. I say this because almost every company in the world produces a stacked bar graph with contributions from major operating segments colored differently - some mitigate the confusion by adding a legend for the colors and putting the percentages in the graph. But others just had a legend, or just had a pie chart, and would find other ways to make it difficult to compare the information to anything else by not necessarily using sales, but gross margin, operating margin, or some company-derived measure of profitability ("adjusted" EBITDA).
It seems like there are really two main categories of mistakes that we can make as data visualizers. The first is stylistic, and the second relates to the substance of the data being displayed. The stylistic errors are clutter/redundancy, incorrect visual dimensions (the overuse of 3-D for 2-D graphs seems to be a major problem), and perceptual uniformity, and "pre-attentivity." Of these, I think the perceptual uniformity is the hardest to get right. Preattentive visual elements can be used to emphasize or de-emphasize something of interest. Stylistic can bleed into substantive through its emphasis or by adding superfluous elements clearly meant to persuade rather than to simply elucidate.
The substantive errors appear to be ones of omission and distortion (the "important to live in a democracy" is a good example).