the subtle differences you're seeing can also be chalked up to the RTA method, as you are probably already aware. Obtaining a legitimate apples to apples with RTA software is very tough. That's why I'd stick with impulse if that's what you're trying to achieve. Then you can gate the responses to be time matched to a 't', and in relation to the distance the speaker is so you're not catching all the other stuff going on. moving the mic just a touch can result in different results; and measuring just a touch longer in time will have the same effect in our car environment. just some input that you probably don't care about at this point.
I think your aiming is fine. I would, too, not aim them across the dash. Cross firing opens up a bag of worms and while beaming isn't until around the 3khz, that's assuming a 1/2 wave pattern. You have to worry about the radiation of the speaker itself but also with respect to how it interacts with your tweeter. Some home audio guys will use 1/4 calculation to determine the maximum delta between tweeter and mid for this reason. There's things about the radiation of a speaker that I can never quantify with words, fully. And when I try it sounds like hocus pocus. Suffice it to say, unless the design calls for it or the speaker has poor behavior on-axis within 30deg, I'd not fire them across the dash at 60-90deg off-axis. You'd be opening up a can of worms that is going to be tougher to correct.