I just ran a test and my readings agree with yours, and on analysis I agree that the issue is in the conversion from voltage to temperature.
I based the calculations off of this document:
Now looking at the implementation and found the issue. The conversion leaves off the top few terms of the polynomial, and the error agrees with your measurements almost exactly. You can see the calculation here:
So thank you for this, I’ll put the correction in the next beta release. I’d like to send you a coupon code as a thank you, can I use the email address you commented with?