A KeyWords cluster, like a WordList cluster, represents two or more words which are found repeatedly near each other. However, a KeyWords cluster only uses key words.
A screenshot will help make things clearer.
These are clusters computed using the Bible as source text. Each of the words here is "key" by comparison to a reference corpus; the clusters show cases where these KWs occur within the current collocation horizons. The [...] brackets represent cases where the KWs are not found together, e.g in come [.] pass there is one dot because the repeated occurrences are come to pass.
See also: Plot calculation.