there's something I don't understand, not coming from a mapping background: sixth decimal place of latitude/longitude is already down to about 10cm precision, so using two 8:24 fixed point would give you comparable precision with half the encoding length than this format (using two fixed points dword encoded straight in base 32). why these encoding prefer to use space partitioning more than encoded coordinates?
... latitude and longitude provide an exact location, are used internally by GPS and satellite navigation devices, and are sometimes printed on paper maps. However they are rarely seen on city or street maps and are difficult for people to use. They consist of long and complicated numbers, have different ranges (-90 to 90 vs. -180 to 180) and need to be used in a specific order. (To express a reasonably precise position requires between 14 and 18 characters.)
A system of encoding location information into a short and an easy to use code would solve these problems.
It goes on to say:
Easy to use: The codes must be short enough to be remembered and used. This means that they need to be shorter than latitude and longitude and about the same length as a postal code or telephone number. The symbols used to make up the code should not include characters that can be easily confused (e.g., 1 and I, 8 and B, 2 and Z, 5 and S etc.)
Does this not answer your question?