I am really not qualified to say much about syntax, or NT Greek in general really. I too am a beginner. And study in Hebrew is not yet really begun. But for what it may be worth here is my current thinking on this sort of thing.
Yes I tried the same sort of thing. And I think at about "within 5" it was getting those sort of numbers. My thinking on syntax versus non-syntax based construct searches is that the additional support of syntax should help with reducing false cases, particularly false positives. But manual vetting of results will be required.
There are also mechanical problems if you like - how do I translate a classification from say Wallace, into a query against Acc syntax according to Fabri. Marco has done a bunch of examples like this one to show people how to do this sort of thing, but I've only played with a couple. I tried as you did to reproduce his example in one case (ὸτι I think) and ran into similar difficulties. In the long run I want to be able to take a construction described in some grammar, such Wallace, BDF etc. and then construct the applicable query and run it. I had hoped to initially familiarize myself with Acc syntax by doing what you did. I wanted this partly to gain confidence in the s/w. Now I think I'll have to approach this by comparison of the results from Acc with evaluation of the returned passages by manual application of the rules supplied by various grammars. This I would expect to help my own understanding of syntax greatly.
There are many questions I have on these topics but I haven't formulated them properly yet. I want to do an in-depth study of some aspect of syntax as an example to help clarify them. But overall now, I've become less concerned with matching the example result counts, for reasons exposed in this thread, and more concerned with validating the results themselves. That's more time consuming of course but probably more educational too.
But I'm not quite ready for this yet. I'm trying to finish up Mounce's reader. Another couple of weeks I hope. Once that's done I have Wallace's and Decker's. So the syntax module will come up at some point, hopefully soon, but probably not before the new year.