In my case it depends on the experiment I’m doing. I like to have protocols as blocks that are always done as one.
A western blot, for example, would be one protocol, as I always try to do it from lysis to revealing. An immunostaining will also be only one protocol. I have 3 different immunostaining protocols depending on the kind of sample I have.
A cloning, in the other hand is split in few protocols: DNA digestion, dephosphorylation, PCR, electrophoresis, ligation and transformation. Depending which cloning I do, I just add the protocols needed. Otherwise I’d need to do a protocol for each kind of cloning, with all of them sharing certain steps (transformation) and with variations on the intermediate steps.
This way, I have a lot of flexibility to put various protocols as if they were pieces of a puzzle, or just one protocol if it always follows the same number of steps.