Form a practical standpoint any series of measurements is only as accurate as the least accurate individual measurements. That is if you have a measurement of a rectangle with one pair of sides at 10.05 and the other at 6, your answer can only be 60. If the first measurement is 6.0 your answer is 60.3, never 60.30. In the real world any thing that is being measured by inference, more than three significant figures is a opium dream.

The number of significant figures that are appropriate should depend on the margin of error of whatever the analysis is. With a very tight margin of error I could see four or five significant figures or even more. Obviously, not the situation in the crude oil production world. Might have relevance in the spectrographic analysis world or the electron microscopy world.

Trouble is, when looking up data, this information is almost never reported with the data, so we are left to guess, and guessing on the fuzzy side is usually the best policy.