[1.0.5] Autopilot still needs tuning in fairly typical scenarios

I try to avoid "X is broken" type reports as I know they're of limited utility unless it's something obvious, but I'm not sure what context would be helpful here.

I have continued to notice that the autopilot struggles in situations that it very much should be able to cope with. I have observed the following:

1. The autopilot is inconsistent with regards to maximum bank angle and turning rate:

   1. When high and fast it sometimes behaves the way I'd expect it to when low and slow -- faster banks to more aggressive bank angles -- but at other times it's the opposite and it will lurch suddenly to the maximum bank angle when high and fast and whip around when changing heading, etc. I assume the actual TriStar autopilot had logic built in to make inputs less drastic at cruise altitude but it seems like if they're represented in your autopilot that they're getting mixed up in different regimes.

   2. When intercepting VOR radials or a Localizer/ILS at 30 degrees or less, the autopilot almost always struggles to begin the turn soon enough and turns too slowly to intercept without overshooting. I am not aware of any TriStar quirks that would make this be the case but I am always willing to learn.

2. The autopilot continues to struggle with maintaining course in a crosswind both when tracking VORs and tracking the INS. I am not even talking about significant crosswinds like a 100+ knot wind at a 90 degree angle, even something modest like a quartering tailwind at around 35 knots it will struggle to track the radial.

3. If intercepting a VOR/LOC/ILS radial in heading mode, when the HSI shows even a half dot deflection and you switch to the VOR/LOC/ILS mode, the autopilot will often quickly bank "away" to intercept the radial more aggressively as if you were farther away, and then immediately but more gradually banks back to turn into the radial to track it, resulting in an overshoot.