Originally Posted by davidfor
Calibre is reading the annotations from the devices internal database. If the device can't display them, then calibre won't find them.
Not directly related to the quote but I think somehow interesting. If one highlights or adds notes to passages which are situated in a range that is referred to in the navigation document by a fragment identifier (e.g. part_0002.xhtml#aliquam) the ContendID in the Bookmark table sometimes contains the fragment identifier, and sometimes it does not. If it does contain then chances are that the device does not display the concerning passages in the main text as highlighted or only sporadically displays them so. They are however listed in the "Annotations" list. So sometimes one can have the impression that a note/highlight is lost even it is actually there. Removing the fragment identifier from the ContendID makes those highlights/annotation behave normally. I did not notice side effects by this, apart from the chapter headings in the annotation list, up to now, but cannot exclude them.
I don't know by what it is triggered that the fragment identifier gets included in the ContentID. I suspect it might be related to how one gets to the passage, for instance whether by simple turning pages, using the TOC, by a word search. (Edit [some days later]: I understand it better now, cf. this post.)
Note: My observation refers to KT.