Where's where?

Time for another Random Linguistic Observation (henceforth RLO). I'm calling this one The Mystery of the Missing Location. I stumbled upon it in the main Dutch reference grammar, where it's referred to as 'to know with the meaning of to known where'. Here's an example:

1. 1. 1. Ik
      2. I
   2. 1. weet
      2. know
   3. 1. Jan
      2. John
   4. 1. wonen.
      2. live
2. "I know where John lives."

Note the discrepancy between the gloss and the translation: the otherwise obligatory locative complement of wonen 'to live' is missing, yet the sentence is perfectly grammatical and moreover, it is interpreted as me having knowledge of where John lives. This in and of itself would be enough to grant RLO-status to this construction, but as it turns out, it has a couple of additional quirks. First, the location disappearance trick only works with complements, not with adjuncts:

2. 1. 1. *
   2. 1. Ik
      2. I
   3. 1. weet
      2. know
   4. 1. Jan
      2. John
   5. 1. slapen.
      2. sleep
3. intended: "I know where John sleeps."

Second, the embedded verb cannot be ditransitive:

3. 1. 1. *
   2. 1. Ik
      2. I
   3. 1. weet
      2. know
   4. 1. Jan
      2. John
   5. 1. het
      2. the
   6. 1. boek
      2. book
   7. 1. leggen.
      2. put
4. intended: "I know where John is putting the book."

4. 1. 1. *
   2. 1. Ik
      2. I
   3. 1. weet
      2. know
   4. 1. Jan
      2. John
   5. 1. zich
      2. self
   6. 1. bevinden.
      2. be.find
5. intended: "I know where John is."

In (3) I'm using the verb leggen 'to put', which requires both a direct object and a locational complement, while (4) features zich bevinden 'to be' (lit. to be-find oneself), an inherently reflexive verb that means the same as zijn 'to be' in its locational sense, as in I am in London (and note that when used in that sense, zijn can occur in the construction under discussion here). Leggen is truly ditransitive, zich bevinden only superficially so (due to its being inherently reflexive), but for the mystery of the missing location, they're all the same, i.e. excluded.

The third and final quirk of the construction is that it is only locative complements that can go missing. Verbs with complements expressing duration or manner do not allow those complements to disappear when selected by weten 'to know':

5. 1. 1. *
   2. 1. Ik
      2. I
   3. 1. weet
      2. know
   4. 1. de
      2. the
   5. 1. film
      2. movie
   6. 1. duren.
      2. last
6. intended: "I know how long the movie lasts."

6. 1. 1. *
   2. 1. Ik
      2. I
   3. 1. weet
      2. know
   4. 1. Jan
      2. John
   5. 1. te
      2. to
   6. 1. werk
      2. work
   7. 1. gaan.
      2. go
7. intended: "I know how John operates."

I must admit I'm quite puzzled by this: I know of no other context that is so specific and selective in its ellipsis properties. If any of you have any ideas, or if the phenomenon occurs in languages other than Dutch, I'd love to hear about it.

1. I'm giving the Belgian Dutch version here. In Netherlandic Dutch, the complement of weten 'to know' would be a to-infinitive, not a bare one.

2. If any of my first-year students are reading this: this makes for an excellent test to distinguish between a locational complement ('voorwerp van plaats') and a locational adjunct ('bijwoordelijke bepaling van plaats').

3. I guess the ill-formedness of (6) could be due to the fixed expression te werk gaan 'operate' (lit. to work go) being ditransitive in some weird sense (cf. the second quirk), but I know of no verb that only has a subject and a manner complement.