I have conducted some statistical analysis on all PMDG aircraft - none of them increase fuel burn 3% per 10 degree ISA increase, and none of them have decreased fuel burn for the reverse.
PMDG have been notified and are looking into the problem.
In the meantime, I hope this might help a couple of people who might be frustrated by not being able to find a good "bias" number. If you can edit the aircraft file, remove any reference to the increase/decreased burn in varying ISA conditions. Or adjust by average on a per flight basis.
Here are some figures for bias to start with that I have obtained by flying the aircraft at different weights in different ISA conditions, and measuring fuel burn (if you're interested in the method, I'll detail that later). All reference the default profiles, which are good profiles:
Tested at CI20
ISA bias -2.47%
ISA bias: -2.03%
Boeing 747-400 RR
Tested at CI50
ISA bias: -2.93%
Tested at CI100
ISA bias: +2.73%
Tested at CI100
ISA bias: +1.28%
Now, if you can edit the aircraft files - I would suggest removing any reference to added/decreased fuel burn for ISA deviation. You can post if you need help with this in this post.
Or - you can look at the average ISA deviation for the flight, and change the bias accordingly, i.e. if the average ISA deviation is +6, you would get 6/10 = 0.6. Multiply by 3: 0.6*3=1.8. Because the ISA deviation was positive, you will burn more fuel. Add the 1.8 to the ISA 0 bias figure. This method is a little cruder, but it has served me well for a long time with some aircraft.
Hope this helps anyone who feels they can't get the right bias for the PMDG aircraft.
Any questions let me know.