You have come across the problem with significant figures. That is, your instrument reads and gives you useful numbers to a certain decimal place. No matter if it's 0.000 or 0.989, it still gives only SF to the 3rd decimal place. I used to work in an analytical lab, and we used to laugh at Sig Figs, because we never used them. The true measure of sig figs is the limit of accuracy of your instrument, and the real measure of accuracy is error analysis (the tolerances on your measurements). Here's how I would handle this:
Take all your numbers, plug them into a linear regression algorithm and then get a number out of it. Round the number to 3 decimal places since your figures were taken with 3 decimal places of accuracy.
You could think of the idea of drawing a line over the significant digits in the zero value, but it still is a problem that you have 4 SFs from the value that's over 1. Personally, I think the whole SF issue is really lame.
If you have an idea of what the tolerances are for your material, then you can use error analysis (i.e., additive and multiplicative error) to determine the final error in your answer, which may be a more fair assesment.