The story is getting a little more complex, and I'm not sure I fully understand what's going on. Or, I simply have too small of an n to be able to say anything widely generalizable. Which is very possible, given that my n is around 20. I've been spoiled by datasets where my n is 100,000+.
First off, about the trust measure. It's a 11 question inventory that loads on two factors. Off the top of my head, I think the two factors are an "emotional trust" (would I discuss work problems with this person/team?) and a "cognitive trust" (does this person deliver). It's an established inventory, and for the purposes of my work, I have to assume that it measures what it says, and the constructs are real. It's also been published, so I can send people copies of it. (And, if I could get you to give a modified version of my survey to your team, I'd love it...)
Now, it turns out the trust numbers aren't as bad as I originally thought. I was mis-coding a few questions, and it ended up skewing the final number down. The initial trust measure is now 2.6 (0 to 4 scale), and the second trust measure is 2.67 (taken several weeks later). This is better, but still not great. The two are also highly correlated (uh... r = .80, p < .001), which is nice. The initial bits of data were not correlated, and I had no good explanation for that.
Tenure with the company is still negatively corrrelated with trust in team mates. (r = -.295, but p = .268, so even the most generous readings of significance say "don't worry about this"). But my new scary bit of data is that the second trust measurement is negatively correlated with the time on the team. AND it's significant: r = -.5, p = .025. I think one team is sort of dragging me down here, but I can't exclude them just because they openly admit they are dysfunctional.
Almost nothing else in this data set reaches significance. I'm trying to figure out if I can get some internal money to pursue this a little further. Unfortunately, in this budget climate, that looks unlikely.