We all go through the safety blanket phase of wanting to see something to sync on, but if you can get blind solving to work reliably for your rig you don't actually have to be pointing at anything.
As Peter says, you can mix and match sync models, within reason.
So you could blind solve to get an initial sync, then once you have one or two points for gross alignment of the pointing model, switch back to dialog based and let the SGP iteratively zoom in on your object.
For an object you're not sure of exact co-ords, just enter best estimate you have from software or elsewhere, then if you can get even a little of the object onto your FOV, use the "point telescope here" functionality to get centred where you want to be.
Then solve manually, and save those co-ords as your master position.
Hope that makes some sense