I’m (apparently) getting false positives, as can be seen here:
Words are being marked as possible spelling mistake, while the equivalent form appears in the list of suggestions for correction. Might that have relation to the encoding of the text I pasted?
Problem solved, I guess. When copying directly from PDF into the stand-alone LT App, somehow this text is not really “plain” text. Words, which contain a link or anchor, somehow retain some symbols or so. This is invisible from outside, but the underlying encoding seems to differ from plain words.