I used fractional indexes for insurance for SD and DD games (+1.5 hilo for SD, +2.5 hilo for DD, rounded to nearest .5). I also made composition-based adjustments. I don't claim this made much difference, but my thought was that, not only is insurance the single most important count-based strategy departure, its index also happens to be right in between two integers.
I suspect that, though fractional insurance indexes are worth little, they are still probably more valuable than some of the soft double and split indexes I learned. Are they worth it? Well, learning more (or more refined) indexes doesn't add much difficulty to counting. Using a more complicated count, however, can be a significant increase in difficulty. So for almost no extra work, you get slightly more than no extra money