The biggest drawbacks to a strategy of highly-specialized centers of production are that losing any one center of production can throw your economic balance way off (especially if your empire has relatively few centers of production, or if you have the bad luck of losing a lot of centers of production that all produce the same output) and that it is highly inefficient to redirect the production of a highly-specialized center of production into an off-type output.
Let's say, for example, that I have an empire of 10 worlds with identical base production. 5 are lab worlds which produce 10 research points (rp), 1 credit (bc), or 1 manufacturing point (mp) per unit of production; 3 are factory worlds which produce 10 mp, 1 bc, or 1 rp per unit of production; and 2 are purse worlds which produce 10 bc, 1 rp, or 1 mp per unit of production. This means that ~90% of my empire's research potential and likely 100% of my empire's actual research output is concentrated in 5 planets, ~80% of my empire's manufacturing potential and probably ~100% of my empire's actual manufacturing output is concentrated in 3 planets, and ~70% of my empire's income potential and probably ~100% of my empire's actual income (excluding tourism and trade; we'll assume that tourism isn't enabled and I have no trading partners) is concentrated in 2 planets. Let's assume that each of my empire's worlds dedicates its entire production to its best output type. Now assume that I somehow manage to lose a purse world. Half of my empire's income just disappeared. If I want to make up just 20% of the lost income by cutting into the research output of my lab worlds, it'll cost me about 36% of my empire's research potential or about 40% of my empire's probable research output (probably a little less in reality due to the newly introduced coercion mechanic).
If I instead had an empire of 10 generalized worlds with identical base production, each capable of generating 5 rp, 3 mp, or 2 bc per unit of production, and I lost one of those worlds, I lose only 10% of the potential research, income, and manufacturing of my empire, rather than ~18% of the research potential or ~27% of the manufacturing potential or ~35% of the income potential, and it'll also cost me less to make up some of the lost output on my other worlds. Say, for example, that my 10 worlds were splitting their production evenly across all three output types. If I lose one of them and decide that I must make up 20% of the lost income from the research budget it'll only cost me 1/135th of my research potential. Of course, this ignores the big issue that my specialized empire gets a maximum of 10 units of output per unit of production while the generalized empire in this example gets no more than 5 (and, if splitting production evenly between outputs, gets only 3.3; coercion 'helps' slightly in that the highly specialized empire might get 'only' 7.5 units of output per unit of production, but it also hurts the generalized empire's flexibility as the generalized empire in this example can only specialize itself to 3.75 units of output per unit of production using fully-specialized slider settings as opposed to 3.3 units of output per unit of production with production evenly split).
In GCIII, the potential maximum output of highly-specialized worlds is sufficiently greater than the potential maximum output of mostly-generalized worlds that the flexibility of a generalized world is an insufficient incentive to generalize. Additionally, if you play on the larger maps or with greater numbers of habitable worlds per empire, it becomes a lot less likely that losing a few planets is going to seriously unbalance your economy, and moreover the more colonies you lose, the more likely it is that you've lost a relatively balanced chunk of your economy - if I have the 10-planet specialized empire described above and I lose 10% of my empire, I'm a lot more likely to lose a big chunk of my empire's income potential than would be the case if I had instead had a 100-planet empire built on the same pattern and lost 10% of that empire. On top of that, the new coercion mechanic isn't really that well set up to encourage generalized worlds; the mechanic is an efficiency penalty on focusing production into a single output, which means that the mechanic actually helps alleviate one of the weaknesses of a specialized world (it's more efficient to generalize the output of a specialized world using slider settings than it used to be) while weakening the main advantage of a generalized world (it's now less efficient to use slider settings to specialize the output of a generalized world than it was), though the mechanic does bring the maximum potential outputs of fully-specialized and fully-generalized worlds closer together.
