Patch and stitch is a technique used for localization in wireless sensor networks. It combines the accuracy of centralized schemes with the computational efficiency of distributed approaches. The network is partitioned into small overlapping patches, localized and finally merged to form a single global map. However, due to noisy distance measurements, ordinary merging may lead to wrong reflection of patches causing large localization error. In this paper, we present a new patch and stitch algorithm which improves the map construction as well as map stitching phase to increase the accuracy of the algorithm. In map construction phase, each node, along with two-hop neighbors builds a local map using Isomap. In stitching phase, maps are stitched to each other in an incremental style. To restrict error propagation (1). a core map with a highest node degree is selected, (2). a robustness criterion that employs flip detect and discard policy is used to ensure a flip resistant stitching and (3). once the absolute core map is formed estimated positions of anchors are replaced by their true positions. Simulation results show that the proposed algorithm enhances the reliability of the location estimates without compromising with the number of nodes localized considerably and thus outperforms the state-of-the-art methods.