I can think of two potential "classes" of reasons for this. (Caveat, I have no special insight to bikesharing, these just seem like naive benefits to avoiding docks)
Real estate for docks is probably not cheap, either monetarily or in terms of "soft capital". (negotiations with city/business owners to allow emplacement). There may also be regulatory hurdles.
Then there's the docks themselves. They likely add a large amount of cap-ex (compared to the bikes themselves) while being "technically" tangential to the actual money-making-value-add of the bikes (so if you can work without them it's pure profit), and are less flexible/repurposable than the bikes themselves, both in terms of meta-strategy (e.g. if a company wanted to launch in a new city, it's a lot easier to just drop bikes on a bunch of corners than to install docks/go through that whole process. Similarly on pulling OUT of a city, much easier to just reclaim bikes) and in allowing better individual bike availability. (docks imply central locations for easy pickup/dropoff, which may not be the reality in a heterogenous city without absurd numbers of docks)
If a lot of this sounds similar to the well-chewed-tech-mantra of "distributed == good, fewer points of failure, more availability" it's because I'm largely pulling from that book. Especially with the sort of data TFA details, this model would probably allow for far more micro-tailoring than having to rely on static docks would.