Censuses in the Desert

Why are the censuses generally, but not always, rounded to the nearest hundred? If it was done with rounding, then all of the numbers should be in hundreds without exception.

Prof. Eli Merzbach quotes some commentators on this topic and suggests himself:

In my opinion both questions can be answered by relying on the following general rules that pertain to fairly large numbers (certainly to numbers greater than 5,000).

1) When the number obtained was in tens (with no units), then it was registered as is and the Torah did not round it.

2) When the number obtained was not in complete tens, it was rounded to the nearest hundred...

Now it is clear why the Torah had to write down the total sums of Israelites in both censuses. Since all the numbers were rounded, one could have had a situation where the grand total obtained would be far off from the actual number in the census. In theory, for the census of the Israelites the deviation could be as great as 588 people... [T]here is a mathematical theorem stating that as the number of figures being summed increases, the deviations resulting from rounding are more likely to offset one another. Actually that is precisely what happened with the census of the Israelites. All the deviations, both upwards and downwards, counterbalanced so that the sum matched the total census taken (of course, to the nearest 50), and therefore it was very important that all these figures and sums be reported in the Torah.