Low Earth orbit (LEO) satellite systems with integrated sensing and communication (ISAC) technology offer a promising approach to provide network access and perform target detection and tracking, particularly for power Internet of Things (PIoT) devices in regions with limited ground station coverage. This paper investigates a partially-connected hybrid precoding design for the rate-splitting multiple access (RSMA)-enabled LEO-ISAC satellite system under electromagnetic interference (EMI) and introduces a novel dual-array ISAC architecture. Specifically, an assisting receiver (AR) is installed spatially separated from the transmitter on the LEO satellite to suppress self-interference and reduce the complexity of time synchronization. Leveraging the statistical channel state information (sCSI) of satellite channels, the target detection and tracking phases are considered for multi-target sensing, referred to as Case I and Case II, where the AR operates without and with prior knowledge of the targets, respectively. To minimize the estimated Cramér-Rao bound (CRB) in both cases, we formulate the hybrid precoding optimization problem and transform it into a tractable form using the Schur complement. A unified optimization framework based on penalty dual decomposition (PDD) is developed to address the problem in both cases, utilizing the weighted minimum mean square error (WMMSE), sequential rank-one constraint relaxation (SROCR), and convex-concave procedure (CCP) methods. Numerical results validate the effectiveness of the proposed algorithms and demonstrate that the RSMA scheme achieves higher sensing accuracy in multi-target sensing than the traditional schemes.