One of the approaches I tried to find acceptable panning and volume settings in my Logic (and GigaStudio) templates was to buy a score book from John Williams (SW episode 4 I believe) and to mockup a few pieces and constantly A/B-ing with the soundtrack while and adjusting levels, panning, EQ and ambience.
This can really help as a startup. After that I kept comparing the loudnesses of individual instruments listening to soundtracks and checking if the relative settings in my template were close to that.
Also make sure that you have a few "required" orchestration and instrumentation books that you will read at least 4 times while trying to imagine what the stuff you read can do for your music and for your setup. Get at least the Rimsky-Korsakow, Adler, Berlioz & Strauss and Forsyth books on orchestration. They all tend to cover and focus on slighlty different aspects of orchestration and instrumentation, and they will also tell you there are no absolute "best practices". Peter Alexander also has some great tutorial stuff (also based on Rimsky-Korsakow) that you can buy in PDF format. I think his RK edition is really invaluable as it adds a lot of current practice to this nearly 100-110 years old book.
Then you will come to a point where you will find that the relative levels in soundtracks can actually vary a lot. Some Gabriel Yared scores for instance have the strings quite up front and prominently, whereas other composers tend to favor the exposure of the other sections. This is where you should (IMO) stick to a global personal template and adjust settings per project that you derive from your template (or reference project file). You can also decide to make submixes of the different instrument groups, render them to audio and make volume and other adjustments in this final mixing stage.
Basically it's all about listening, training your "ears", actively trying to emulate mixes, improving your mixing skills, which all takes quite a while (and will never end). And don't forget to get an acceptable "feel" for the differences between classical recordings/mixes and soundtracks. They can be quite different beasts.
Cheers,
Pete
PS: I also "think" in three groups (from an ambience perspective that is): strings, woodwinds and then brass and percussion. I do, however, have the two typical woodwind seating "rows" as separate input channels, so flutes+oboes/EH and clarinets+bassoons (fagotti). This allows me to have different EQ-ing for these "rows" as well as different early reflections. But in total I have 11 input channels plus a spare channel, so I typically have more separate processing (also dry and wet brass, the strings groups, etc). But it is smart to think in at least 3 "distance" groups.
