Microstepping helps with resolution, but not as much as it initially appears to. The datasheets for many stepper motors specify that the tolerance on the angular position for each step is 5%, for a 200step/rev motor. That means you can say with some confidence that if stopping on a full step, the motor will be accurate to 1/(5/100%)=1/20th of a step. That implies that down to around 1/20th microstepping you'll gain resolution, however when stopping on microsteps the tolerance is worse since the relationship between the reluctance of the rotor, and angular position, is not ideal. It's hard to say exactly how much resolution you can gain. My micro lathe has 4mm pitch ballscrews, geared 2.5:1, so the effective pitch is 1.6mm and with 1600step/rev (1600 is good number to use in general) that's 1um per microstep. If I put an indicator on it and try and measure this, the axes do appear to move about 1um per step, which is promising but bear in mind that's in a static situation. Whilst the motor is spinning continuously it will be less accurate. Arguably the main bonus from microstepping is it means the motors are exciting the system at a higher frequency, so problems with resonance are attenuated.

I can think of very few parts that would benefit from the 2* resolution gain got from using 5mm effective pitch instead of 10mm. There's only one part I've made where I decided to swap my pulleys round to get higher resolution, and that wasn't the Mayan calender. However there are numerous other reasons to use pulleys, so it's still a good idea to plan for the ability to swap them to get the higher resolution. If you stick with the 1605 screws then option 1 (1:2 to get 10mm effective pitch) will work substantially better than 1:1.

Unless you'