I was confused too, and hence decided to spend some time on this; here's what I found.
As per Neil Weiss's book, there're 3 types of grouping of data.
1. Single-valued grouping (uses single-value frequencies and graphed with bar charts)
2. Limit grouping (uses class-intervals - between the lower and upper boundary of each class, and graphed with histograms)
3. Cut-point grouping (uses class-intervals - between the lower boundaries for each successive classes, also graphed with histograms)
The first one is typically used for discrete data. The third is used to represent continuous data, and the second in both cases.
In our case, we're talking about the 3rd case, since ours is continuous data.
That's the reason, it uses cut-point grouping and hence the limits are different - starts from the lower boundary of the first interval, and ends with the lower boundary of the second interval.
I hope all that makes sense...
Please feel free to challenge me on this:) I'm no expert.