So, first of all, typically optocouplers are not meant to be linear measurement devices – as you can see by the fact that you need a graph to represent your current transfer ratio (instead of a single value).
Now, you ask whether there's an optocoupler that has a current transfer for a input current of 8.5 µA.
Probably there's no such device. Remember that this current is just used to power an LED in the optocoupler – and single-digit microamperes aren't really much! You need to put this into perspective with the thermally-induced noise in the receiving end of the optocoupler.
A simple solution would be to just amplify the current – even if it's just with a single BJT – and use that increased current. Sure, the nonlinearity of your transistor makes calibration of the system necessary, but it's already necessary because of the non-constant current transfer ratio!
Or, you just use an isolated ADC (e.g. AMC1035) that includes optical isolation in the package.
Other than that, you could simply use a galvanically isolated method of measuring current and get rid of the optocoupler; measurement transformers (for AC) or hall-effect based measurement comes to mind.