[lkerlk]

I've read some research that suggests that certain characteristics--namely agreeableness--are more relationship-specific than others.

For example:

http://www.sciencedirect.com/science/article/pii/S0092656616...

That meta-analysis shows that rather-specific variance is greatest for agreeableness, although that is misleading (in an underestimating kind of way) because what you really want to know is the target-rater interaction variance, which is often not studied because it requires a lot of combinations of targets and raters. What I remember (and I might be incorrect) is that in such studies, there's also a lot of relationship-specific variance for agreeableness.

The papers I've read by Vazire ignores this interaction variance and focuses on agreement across traits. The interaction variance is implied, though, by the huge amount of variance not accounted by target or rater.

The takeaway message is that there will be a lot of disagreement for most people about who thinks who is a jerk.