The first thing to understand is that this is a defined unit (actually a centiMorgan, one hundredth of the original invented unit named for the geneticist T H Morgan) not something that is based on a physical standard.
Although it is talked about as though it were a measured length,the cM (also called a mapping unit) is actually a statistical notion. The idea of a "1 in a hundred year flood" is broadly similar in linking a measure of length (water depth) to the probability of its occurrence. In the case of genetics, the claim is that if you work 1 cM along a chromosome, then there is a 1% of chance of encountering a "crossover event".
The complication is that crossovers are not equally likely on all chromosomes so that not all centiMorgans contain the same number of base pairs (another way of measuring its length. That's right, not all cM are the same length.
In many ways this is the perfect feature for selling genetic genealogy information. It looks straight forward (compare your number to this range) but very few clients will have the time or the expertise to investigate its basis and to calculate the maths.
To my mind, the best (but expensive) way to use this is to get data for a range of people whose relationships you know (from traditional study) and then form an intuitive notion of how the cM data aligns with that prior knowledge. Then you can make some semi-informed extrapolations about hypothesised relationships on the basis of the cM data alone.
The Human Genome: A User's Guide may be a suitable starting point for further study.