This...

Quote - ... Have you tried breaking the item up into more than one UV map? If you have a model that has lots of "parts" and you try and fit them all on one map, then detail is definitely gonna suffer ;o). Sometimes you can't just "squish" all those different parts onto one map...lol. They're all gonna be tiny.  Laurie

is truth.

Optimizing your UVMap is so critical everywhere, most of the time, that people do it on instinct.  But, for rendering stills or light animation work, it's not necessary to rape every bite of space you can and, quality can suffer because of it if you're primarily worried about the 2D texture effects.

SO, UVmap out sections separately where greater detail is needed than a single map, with all its gajillions of zones, can afford.

Pull up a V4 UVMap in UVmapper. What do you see?  Multiple maps, that's what.  Why?  Because, the detail required (in a good 2D texture) for a rendering figure is much greater than that required by a online game and memory space is not usually at a critical premium for simply rendering it.  So, V4 has a bunch of x-thousand by x-thousand maps covering her torso, face, legs&arms, etc. which give very good results.  If she was in a game, she'd have one UVMap with everything crammed on it as low as resolution as possible yet still giving decent results.  Not easy to do... there are specialized programs and plugins out there costing big bucks that automagically try to optimize UVMap efficiency.

If you're using that airship (nice, btw) for high detail rendering, try to use comp generated materials for the skin.  That probably requires less exactness of placement than anything and would be easily done.  You can also augment it with a high-res bump map if necessary or just use one with materials generated by the comp.  (Poser's weave node or tile node comes to mind for a good, comp generated bump that takes up very little space.)

But, whenever you need more detail for a 2D texture, always provide room for it on the map, even if you have to make a separate one. Which is no big deal and very easy to do if you need help there.  The computer doesn't care about what a 2D UVMap looks like to us with stacks of UV'd faces all on top of each other.  It doesn't read it in the same way as we see it on the screen.  So, when you open up a high detail object and see stacks and stacks of mapped surfaces/shading domains lying all on top of each other, it may look like garbage but, the computer loves it just the same as one "optimized" UVmap.  It's just a tad bit more memory hungry because of the increased resolution of the maps.