If you have records that bad you probably need to add a little hepastat. If that still doesn't work you need something that can break down a biofilm. In that case you need enzymes with it. I am not certain where to buy enzymes on their own. Pomegranite juice would work but I don't know how problematic rinsing it and the sugars would be - I would test it first. I suspect if its easy to clean off then you could spread it on the disc and let it sit for a few minutes before cleaning with your iso, hepastat, & surfracant mix in the sonic machine.
Along with some IPA and a touch of Ilfotol, I tried Hepastat in my DYI USC but stopped using it. With only IPA, Ilfotol and 2.75 gal of distilled water I measured dissolved solids at 0000ppm. Adding Hepstat takes the total solution to 0028ppm - too much gunk for my comfort level. Tank runs through a 0.35 micron 'polishing' filter during operation.
As to David's comment on heavily soiled records, I believe the trick with a USC is to use a lower then a higher frequency. I run 10 minutes at 37kHz 60% power then 10 minutes at 70-80kHz 80% power. Max power is 330W. All this at 33° C. My USC allows varying time, frequency and power as appropriate. I clean 5-6 records at a time.
I used a Loricraft PRC-3 for over 10 years, mostly with AIVS cleaners (my review) including enzymes. I ran an an AudioDesk for 4 years. I believe you can get a record as clean as it can get with a point-nozzle vacuum and mild agitation to keep particles suspended in fluid, but that approach is very time consuming, messy, low on efficiency, and may take multiple iterations.
Current results are as good or better than I found with a point nozzle vacuum. I use the Ellington test to gauge results.
Ultrasonic frequency may correlate with particle size removal.