You construct a sample in which Whites, Blacks, and Latinos are randomly selected from the U.S. adult population. The composition of your final sample is 25% White, 30% Black, and 30% Latino. You've sampled from all over the country, but the actual composition of the population has a much higher proportion of Whites. This is likely an example of a:

When you use stratified sampling, you divide the total population or universe into strata (non-overlapping homogeneous groups). Then you draw a simple random sample (SRS) from each stratum (singular of strata).

You must make sure that your strata does not overlap, e.g. Latinos are not a race, but a culture and they come in all colors (whites, native indigenous, blacks, even Asians). That is why sometimes the term non-Hispanic whites is used.

Another reason why the sampling error might have occurred is that not all Caucasians are white, e.g. Arabs, people from northern Africa and most middle eastern Jews are considered Caucasians but their skin is tanned and it can be as dark as the skin of a Native American.

The US census only recognizes 5 races:

Whites (includes Hispanic whites and other Caucasians)
African Americans
American Indian or Alaska Native
Asian
Native Hawaiian or Other Pacific Islander
the sixth category is two or more races

As you can see, there is a lot of room for mistakes when you try to categorize people into races.