[The following is addressed to whomever is reading this thread, not specifically the OP]:
If the SHT31 read 99% or 100% recently then, perhaps, the 96% reading is valid. But if you notice, as I and many others have, that the ceiling for the sensor goes from, let's say, 98% upon installation but over time drops to 96%, then you can assume that the sensor isn't going to read properly during fog. I've got three 31s. The humidity maxes are: 97% (newest), 96% (2nd oldest), 95% (oldest).
Kestrel produces their own in-house sensor. It's hard to imagine why Davis would pay an engineering team to do the same. If Davis were selling these instruments primarily to people who knew better and it were a matter of life and death or they were in a higher market tier, then we might see a different response. Kestrel, on the other hand, sells its equipment to the military, firefighters, sports teams (heat stress on the field), storm chasers and other critical industries.
Based on my own tests involving visits to local airports and other professional weather stations, I can attest to their accuracy. I know another forum member, Ken, who has a masters in met, who has tinkered with them extensively and will attest to their accuracy. They are good, reliable devices when used properly.