How high code coverage can indirectly measure quality?

In many books, the following statement is quoted:

This analysis (--code coverage) gives a quantitative measure of the coverage, which is an indirect measure of the quality.

The way I see it, even 100% of statement, branch and even conditional coverage cannot guarantee the code is quality - the tests can be incorrect, the logic (in both source code and tests) might be misunderstood and the coverage might be satisfied artificially. I am not a developer so maybe I am missing something - how can it be seen as a indirect measure of quality?