When manually calculating grades, teacher average each marking period and then would average those averages to find the final average. While that was the simplest way to do it, averaging averages compounds rounding errors and gives undo weight to marking periods with few grades.

In contrast, Candle averages each marking period and also averages the whole year to find the final overall average. This give a much more accurate picture of what the student has done and will often be different from an average of the averages.

Here's a contrive example for demonstrative purposes. Suppose a student got these grades in a subject

Period 1 - Avg: 94 Grades: 95 94 92 96 94

Period 2 - Avg: 91 Grades: 90 92 91

Period 3 - Avg: 80 Grades: 80

Period 4 - Avg: 92 Grades: 90 92 91 95

Average of averages: 89

Average of whole year: 92

As you can see, because period 3 had fewer grades, AND because that grade was poor, averaging the averages gives undo weight to that one grade. Averaging the whole year usually gives you a better picture of how the student actually performed. 

An exception to this can occur when you use true weighted averaging (the default method) and only have a few grades of a particular type. For instance, the default weight for quizzes is 17% of the average. If you only gave one quiz during the entire year, that one grade will have an unduly large effect on the final average since it counts for 17% of the total grade. It is important to consider how many of each type of grade you enter and set the weights of those assignment types accordingly.