If you need a spacer to adjust chain alignment, you can get bottom bracket spacers (not lockrings) in 1/2, 1, 2, and 4 mm thicknesses. They're cheap (about a dollar) and aluminum.

The problem with using a spacer there is that any other wheel will also need a spacer, plus the spacer acts like a washer -- it creates "slipperiness" between the cog and the face of the hub (not the threading but the perpendicular surface that the cog butts against). This makes it easier to disengage the cog. Plus, as Nullius said, you can run out of lockring threading.

I'd actually recommend moving the bottom bracket slightly if the cog needs spacing. If you are spacing the cog, then your bottom bracket needs to move slightly to the left. There are bottom brackets that can be adjusted in both directions but if you have one that simply cranks down tight on the right side, you'll want to ask your local shop to face the bottom bracket shell slightly.

Does your cog have good alignment? Noise can be a matter of cog shape and quality, not just alignment. If you're using a Surly cog you will have more noise. I like EAI and Dura Ace cogs the best, but even among EAI cogs there's a lot of difference in noise and smoothness between the standard black one and the stainless one with the polished edges. Check for chain alignment on the cog. Let that be the deciding factor on how you space anything. And be sure you're using a standard dimension cog (like a Dura Ace, EAI, or Phil cog).

There are a number of single-speed freewheels out there that have different spacing from cogs. That may be one source of your problem.

In short, you could be having problems with an inexpensive cog, a non-standard spaced freewheel, or simply be hearing noise because of the cog shape rather than real mis-alignment. Two millimeters of chain alignment is rather hard to detect and can create a bit of a difference in noise and smoothness but only if it makes the difference between being on the edge or simply out of alignment. Your measurement from the flange to the cog may not really be that relevant to actual performance.