Clearly one disparity is the generalized torque specs are based on an assumption about what is being threaded.
Threading into a casting, I've got to think the cast material and shape is going to be a big factor.
When we developed torque specs, we would take all of the parts involved in the joint and torque at least ten "sets/samples" to failure with a DC electric wrench that monitored torque vs angle and the fastener engineer would use that sample size and information to determine the torque specs for that particular joint... I was involved in this and was right there doing the testing on my parts with him...
We would look at the graphs and determine where the ramp bent over 10%...