we've been hashing this around at the office today, and no one is coming up with a reason not to do this as a contact neg-- just putting the 4x5 film and 120 positive together with a glass over them and flash that with white light.
This is a much better option, IMO (as I said above, though I did not say why I thought so), for more than one reason.
First, you get the minimum possible generation loss (aside from scanning). No enlargement, no lens introduced into the equation, etc.
Next, you don't have to focus such a tiny image. Less chance of user-introduced (focus) and equipment trouble (misaligned enlarger, shaking enlarger) into the process.
Third, you can raise your enlarger higher to get finer control.
Additionally, you want an exposure long enough to give you the fineness of control that you need to get the ideal exposure (1/3 stops are fine). I would not fear reciprocity failure too much. For one thing, you can get a film that has great reciprocity maintenance during long exposures. Even with reciprocity loss, you sill want longer exposures, however, IMO. The exposures I was getting for my masks on Tri-X (EI 320) using my Omega B22XL (all the way up at f/16, and no filtration because I was doing black and white to black and white) were in the range of normal print times. 16 seconds, I believe...and this film loses reciprocity like a muttha$*@$.
Also, remember that you enlarger does not cast white light (if it is using a tungsten bulb). It casts reddish-orange light. You need to add cool filtration (approx. 80A + 82A) to get the light to be white.
Another thing you can do, if you feel so inclined, is to use a flash to make the exposure. I did this one time with litho film using a 500Ws Dynalite studio flash, a small softbox and ND gels because my enlarging lens was loaned. If you have one that can be manually set in third stops, that is great. ND gels may help too. You can use the guide number of the flash or a flash meter to figure a first-try at an exposure. Put the sandwich in the bottom of a black box to prevent the scattered light from hitting.
Any light will work, as long as you can somehow rig it to provide 1/3-stop precision.