We study here one-dimensional model of aggregation and fragmentation of clusters of particles obeying the stochastic discrete-time kinetics of the generalized Totally Asymmetric Simple Exclusion Process (gTASEP) on open chains. Isolated particles and the first particle of a cluster of particles hop one site forward with probability p;when the first particle of a cluster hops, the remaining particles of the same cluster may hop with a modified probability p m , modelling a special kinematic interaction between neighboring particles, or remain in place with probability 1 − p m . The model contains as special cases the TASEP with parallel update (p m = 0) and with sequential backward-ordered update (p m = p). These cases have been exactly solved for the stationary states and their properties thoroughly studied. The limiting case of p m = 1, which corresponds to irreversible aggregation, has been recently studied too. Its phase diagram in the plane of injection (α) and ejection (β) probabilities was found to have a different topology.Here we focus on the stationary properties of the gTASEP in the generic case of attraction p < p m < 1 when aggregation-fragmentation of clusters occurs. We find that the topology of the phase diagram at p m = 1 changes sharply to the one corresponding to p m = p as soon as p m becomes less than 1. Then a maximum current phase appears in the square domain α c (p, p m ) ≤ α ≤ 1 and β c (p, p m ) ≤ β ≤ 1, where α c (p, p m ) = β c (p, p m ) ≡ σ c (p, p m ) are parameter-dependent injection/ejection critical values. The properties of the phase transitions between the three stationary phases at p < p m < 1 are assessed by computer simulations and random walk theory.