I understand the explanation, but your final statement is still wrong. If you're doing "serious date calculations" for any modern organization, and you are asked to make a graph of 20th century data by decade, you will be expected to use the standard decade divisions, i.e. 190_, 191_, 192_. If you include data from December 2000, and get different results from your colleagues, you have failed. The last decade of the 20th century was Jan. 1990 - Dec 1999, according to everybody except the most pedantic of history majors.

And the idea that you could actually do serious date calculations that count backwards all the way to AD 1 while maintaining a precision of 1 year is also ridiculous. There are error bars around the year numbers from that long ago of something like 2-5 years.

I understand the basis for the [1901-2000] definition. However, I really don't think its defense outweighs its huge downsides, especially the split-decade problem.