Very interesting Bob. I reread the bit of Wallace on the point and he defines two cases : RP and ant 1) differing in respect of gender and 2) agreeing in case presumably where the natural case of the RP be different. Constructing a query for the first is easier than the second but reveals a couple of concerns. First scope : its unclear if the antecedent is always going to be in the same verse or not, but doing a verse search is at least a place to start. Secondly the syntax module is not yet complete. Thirdly and perhaps most concerning is an example I found which seems to demonstrate a shortcoming probably in the construct capabilities. Attached is an image of the simplest query I could think of for the first case.
It brings up 361 hits. Hmmm... that seems a lot ... given your friend's estimate and the incomplete syntax module.
Pulling up the syntax module next to one example and looking at it for a long time I found something I do not understand. Here is the snap of that window.
The passage is Matt 3:17. You'll notice that the RP is highlighted and that before it two missing words are indicated. One of those (the second) is an RP with φωνη as the antecedent. But the antecedent to the highlighted ᾧ should be υιοσ which is not highlighted. And you can see that they are in fact related because they are both tagged with the same antecedent id 3. But of course neither ᾧ nor υιοσ should be highlighted because they are the same gender. So what is going on ? My initial guess is that the antecedents returned by the construct are not known to be the antecedent of the RP. In fact one is not an antecedent at all. So I thought, and finally it dawned on me that an Agree on the antecedent id was what I needed to add. No dice though - you cannot specify agreement of that kind.
I would be really interested in other input on this because this looks like a problem with construct searches not being able to express a required form of agreement. Of course I could also be all wet on this in which case I love to hear how I should do this.
There is one further fly in the ointment. The Getting Started with Syntax Guide mentions antecedent ids and gives this description : "3. Antecedents — To assist in showing syntactical relationships, antecedent identifier
labels are sometimes added to show relationships between a word and its
antecedent. A null identifier may also be combined with an antecedent identifier."
The bold is mine - if they are only sometimes present I do not know how to do an effective search for a construction involving a word and its antecedent, even if I have the Agree (ant. id) function.
So Bob, I don't know the answer to your question and I've no doubt wasted a good deal of your time reading about my confusion.