Question though… I couldn't find (or perhaps missed) which LDs (DC, faces, NTSC or PAL) you are using to accomplish this transfer.
I used the NTSC Definitive Collection LDs in my transfer and the one thing I was able to accomplish that no other group seems to have bothered with was PERFECT deinterlacing back to the 24FPS source. After fiddling with filter after filter that attempted to perform IVTC, I finally spent the time to MANUALLY determined the PATTERN of the interlacing and exactly WHERE it reset/errored.
After mapping out the 3:2 errors and accounting for them mathematically I made an avisynth script that actually disassembled the fields into 24fps frames rather than artificially (via filter AI.) Now, I am not positive about the “faces” LDs, but the DC LDs are EXTREEMLY close to being a perfect 3:2 pull down. Buy that I mean, except for one or two changes in the pattern you can actually apply a mathematic deinterlacing and get all the actual frames back flawlessly.
I’d love to know what you are doing for IVTC and if my frame counts could be put to use to allow your work to be perfectly deinterlaced I’d LOVE to participate!
Zottig