Wireless Sensor Networks (WSNs) have been envisioned for mission-critical applications such as critical infrastructure protection where reliable data delivery, goodput, and energy efficiency are of paramount importance. Due to the resource constraints in WSN devices, several transport protocols such as RMST, PSFQ, DTC and DTSN have been designed to leverage intermediate caching in order to avoid the costly end-to-end retransmissions inherent to traditional transport protocols such as TCP. Transmission window size adaptation, acknowledgment semantics, and loss recovery are important components in the design of the transport protocols. TCP uses the additive-increase multiplicative-decrease (AIMD) algorithm and cumulative ACK mechanism. More recent works such as DTPA have shown that TCP's AIMD scheme leads to inefficient performance in wireless networks. DTPA uses a fixed-sized transmission window based on the bandwidth-delay product (BDP) of the path but retains the end-to-end semantics. However, with caching based protocols, there is a need to revisit the transmission window size optimization, since the latter has a strong impact on the effectiveness of the cache. In this paper, we provide two optimization schemes namely, an enhanced 0(l)-time complexity NACK-based repair mechanism and the optimal transmission window for DTSN. Incorporating these optimizations, we implemented an enhanced DTSN protocol (denoted as DTSN+) and compared its performance with TCP and DTPA. We show that the optimal transmission window of DTSN+ is dependent on the average cache size at the intermediate nodes. Our results show that DTSN+ will in general significantly outperform both TCP and DTPA in terms of goodput and energy efficiency.