Oil drop experiment--How was the result so accurate?

I remember while learning about Millikan's oil drop experiment and being pretty skeptical about the setup. I know that there is a lot of controversy regarding manipulation of data, but the fact is; he still got a pretty close answer. I mean, we can get the correct answer by manipulation of data if we know the answer already. If we don't all we can do is 'reduce' the error box (which Millikan purportedly did). But I'm surprised that he still got an answer close to 1% of the actual value. I personally would have expected an error of maybe even an order of magnitude for these reasons:

• Viscosity of air: The viscosity is dependent upon humidity and other factors. I know he did use the wrong value here.
• $g$: How was this known to the accuracy required for this experiment? It varies across the Earth, and besides, could it be measured to whatever accuracy using the equipment available at that time?
• Velocity of the oil drop: I don't know how fast the oil drops fell, but to me, terminal velocity looks pretty fast. How did he accurately measure the velocity? A micrometer eyepiece takes time to focus, and more time to take a reading. And anyways micrometer scales aren't that precise, are they?
• Potential difference: How did he accurately know this?
• "Selecting one drop": How was this done? If all drops had a charge as a small integral multiple of q, there should be more drops with the same charge which behave the same. How can you get rid of those?
• Densities of air/oil: Air density varies. Oil density may not be known accurately.
• Electric field: Between plates, E is not exactly constant or equal to $V/d$ There are fringing effects which may alter the value of E. IMO since the distance between the plates is comparable to the area (it look like it), fringing effects are pretty significant.
• Stokes law is only applicable for perfectly spherical bodies. Is a falling oil drop perfectly spherical?

Basically I'm skeptical of how one can get such a tiny number with 1% accuracy with all those uncertainities. Could someone convince me how he did get it so accurately? Were there any measures he took to minimise these errors (I can't find any)?

• there has been controversy in his data set selection...en.wikipedia.org/wiki/… Feb 28, 2012 at 5:41
• @VineetMenon I know that, I've mentioned it. I don't care about his reported error box, I'm just skeptical how he got such a close answer. He couldn't have chosen the data set to match the value of e as the value of e was unknown at that time. Feb 28, 2012 at 6:18