Are any of these from Google Books? I've noticed they introduce <a> tags indiscriminately (I think it is to be able to switch from Flowing Text to Paged (PDF) mode without losing reading position). That might be throwing off the word selection when you happen to select one of these words.
Here's an example:
Quote:
|
unemploy<a id="ORIG-GBS.PA149.w.1.0.0"/><span class="gtxt_body" id="para.161.1.0.box.225.244.1351.804.q.60"><a id="GBS.PA148.w.3.0.0"/><a id="GBS.PA149"/><a id="GBS.PA149.w.0.0.0"/><a id="GBS.PA149.w.1.0.0"/>ment which
|
You could crack open one of your problem books with calibre ePub editor to verify this.
Perhaps there is a plugin that can clean these things out. Or if not, perhaps someone should write one. Or maybe you can filter them out with ePub to ePub conversion.