Dana,
<<Is there a difference between having the threads in the bellcrank verses having them within the rod end bearing?>>
Yes indeed. Consider your bolt as a round section cantilever beam fixed at one end and loaded at the other, which is an accurate representation of the problem. The highest material stress in such a beam is found at the support (fixed) end; the equation is (load x length)/section modulus. Section modulus = moment of inertia/distance from the neutral axis of the beam to the most extreme fiber. The MOI of a solid round beam is 0.049 x d^4.
Our trusty AN3 bolt has a "d" of 0.1875" and thus a radius (distance to extreme fiber) of 0.09375" at the shank end. At the other end the thread root diameter is only about 0.160" and radius is 0.080". Assume a beam length of 1/2" and a servo load of 20 lbs, then run the numbers. You'll find that the material stress is about 15,500 psi if you locate the shank in the bellcrank, but almost 25,000 psi if you locate the threads in the bellcrank. You pick up a 60% safety margin just by installing the bolt the right way.
The above ignores fatigue and stress concentration. Fatigue failure is directly related to the number, amplitude and type of load cycles. In this application you have fully reversing load cycles, which are the worst kind. As for stress concentration, AN bolts have rolled threads to reduce the notch effect, but it is common to have a bolt with a less than perfect root thread. A sharp V root thread (rather than a nice rolled U) is a "break here" sign; the base of the V is exactly where you find the 25,000 psi material stress. This is particularily true if you managed to locate that root thread flush with port side surface of your bellcrank.
Is it safe with the threaded end in the bellcrank? Maybe. AN bolts are 125,000 psi material, but stress concentration, number of load cycles, and the amplitude of the load cycles all conspire to chop the margin by many multiples. Why take chances? Whole lot easier to remember that neither shear nor bending loads should be applied to bolt threads.
Usual caveat: I am not a professional engineer, and even if I were you should check my math <g>
Dan