The short analysis:
First: Discard the data from August 1. The data says that you had the luxury of a 27 hour day. Obviously the data is spurious!!
Second: Recognise that the Telkom data is in Bytes and that the Tiscali data is in Kilo Bytes calculated at Bytes / 1024.
So, take the Tiscali running total as at the 11th (start date) and subtract the transfer of the 1st.
552 985 - 187 088 = 365 897 KBytes
Add the Telkom daily combined amounts for the 2nd to the 11th
and get 347 679 003 Bytes. Divide by 1024 and get 365 897 KBytes.
Now 365 897 KBytes = 365 897 KBytes
THEY DO ADD UP!
Understadably the measurements are not synchronised. The difference of 2 055 KBytes on the 3rd is exactly matched by a balancing difference on the 4th. Similarly the difference on the 7th is balanced by the difference on the 8th. Small differences in the data can be attributed to rounding errors.