DT finds string in an img's data-url – intentionally?

If I search for the string “isbn”, DT returns three documents, two of which contain the string (one MD, one PDF).

The third document is a formatted note consisting of an img element with the image in a data-URL. And this data does contain the string, like so:
(I added ~~ around it to make it easier recognizable).

Does it really make sense to return matches from serialized image data?

In my opinion, it would be more sensible to index only the text nodes of a formatted note’s DOM.

Of course not and DEVONthink shouldn’t, only the ALT/TITLE attributes of images are indexed. A copy of the document would be useful. But please use the prefix text: first to ensure that the indexed text actually caused the hit.

Same result with text: prefix.

The compressed document is 14 MB, I’ll send you and iCloud link to it.