This is coming a little later than anticipated but life happens. First of all, scanning negatives for comparison on the web stinks, there are too many variables to create meaningful comparisons. Prints would tell us much more, and if I have time in the next few months I'll try a few of these negatives.
This was a speed test shot in my back yard with Ilford SFX 200 film with; no filter, #99, #29, 680nm, 720nm filters. The #99 is a B+W filter purchased from B&H as a supposed infrared filter, but it looks like a slightly deeper version of an 85b, oh well. With that in mind I scanned each of these strips with all choices turned off and let the Epson scan software choose the black, middle and white points. The images from the higher ISO's have more detail in them than these scans suggest.
Metering was accomplished with a reflected reading off of a gray card with a Gossen luna Pro through the beam splitter attachment set to 7.5
o. I might have used my Spotmatic's internal meter but it doesn't go any lower than 20 and since part of this test was to look for IR effect, I stuck with the handheld meter. Any ISO's with a (-) sign mean that many stops below the slowest speed on the Luna Pro, (.8).
Processing was done in HC-110 at 1:120 dilution, 23
o C., for 14 minutes. Initial agitation was 45 sec. and 5 seconds every three minutes there after--twirls not inversions. The idea was to control highlights while building up shadow. Even though the negatives and scans don't look it, the roll is underdeveloped. But most of the frames are actually overexposed. I got better response through the #29, 680nm and 720nm filters than I was expecting. All evaluations of the negatives are by eyeball and therefor subjective. The ISO ratings go higher to lower from left to right.
Strip #1 was exposed with no filter at iso; 320, 160, 80, 40 & 20. ISO 80 gives what looks like a Zone I exposure just off dead black, but there is little usable highlight information in the neg. even though the scene had plenty.
Strip #2 was exposed with the #99 filter at iso; 160, 80, 40, 20 & 10. ISO 40 looks like the speed point, same comment about highlights.
Strip #3 was exposed with a #29 deep red filter at iso; 40, 10, 2.5, 1.2 & .6. ISO 10 looks like the speed point but it might be as much as 20 with proper development.
Strip #4 was exposed through the "inexpensive" deep deep red 680nm filter at iso; 25, 6, 1.6, -1 & -3. ISO 6 has good shadow detail, same comment as strip #3 above, iso 12 might work.
Strip #5 was exposed through a 720nm filter (also "inexpensive") at iso; 10, 2.5, -1, -3 & -4. The deep blacks in iso 2.5 are almost too light, 5 would have been a better speed. And iso 10 might work with proper development.
I think the 680nm filter has promise as a compromise when trying for IR effect without totally loosing shadow detail. I'll probably run another test to ballpark a development time, then give it a whirl in the real world. Here are individual scans of my pick for each filter.
no filter

#99

#29

680nm

720nm
Again, the scans don't really tell the whole story.
Steve