The behavior I expect is that omission is used any time the text is truncated, regardless of how the end point of the truncation is determined.
#40 had a fix for the omission missing when break_token is used, but was closed with no discussion for some reason.