Not quite right. The Census data counts wealthy people but "top-codes" high income persons. This means it does not report their actual income, but a fixed floor amount (I forget what it is). The rationale is to prevent individuals from being identified in publicly-available data. So the people are counted, and assorted techniques are available to researchers to impute their incomes for purposes of analysis. I've never seen anybody object to these imputation procedures on principle, so there is no real political issue here about biases in the reporting of the distribution of income or wealth.
There is also the Survey of Consumer Finances, which focuses on wealth (and wealthy people), and the IRS income tax data. Both of these "oversample" high income persons to improve the accuracy of the sample.
Bottom line is that there is enough data available to do analysis and make waves, though there is always room for improvement. For examples, see Ed Wolff's book on the distribution of wealth (Top Heavy), or the next State of Working America from EPI.
MBS