Designing points systems for wargames is of course a lesson in pain and futility.
A lot of factors are at play: Different things combine with each other in unpredictable ways and you have to constantly balance between keeping things simple (and often more broken) or less broken (and more complex).
Then you have issues like the gaming table (an open table means long range weapons are more valuable f.x.) and even the gaming group (if nobody in a group plays close combat oriented armies, then they'll find a different focus).
My usual tradition as I work on a points system is to put together some approximate values (for StarStrike, I am using an actual formula), then go backwards and add up some of the test battles that have been played out, and see how they compare.
Currently,about a 10-25% variance in points across a few tests, but then... does that mean the formula is off or that the battles were?
In the end, at some point, you just have say "it's good enough" and go with it.
Of course, a benefit of PDF publishing is that going back and adjusting the values is quite easy.