Knowledge Graphs (KGs) provide structured representations of knowledge, where entities are modeled as nodes and their relationships are captured through typed edges. Knowledge Graph Question Answering (KGQA) focuses on interpreting natural language queries and retrieving accurate answers by leveraging the information encoded in these graphs. An advanced variant, referred to as multi-answer KGQA, requires more complex reasoning across multiple connections within the graph to deliver all relevant responses. A major challenge in this context is the incompleteness of KGs, which becomes particularly problematic when queries demand multiple answers. To alleviate this issue, recent research has investigated the integration of external textual resources as complementary evidence; however, such resources are not always readily available. In parallel, KG completion techniques have been developed to infer missing links and improve graph coverage. Despite these advances, their application within multi-answer KGQA remains underexplored. In this paper, we propose a novel system specifically designed to enable effective multi-answer KGQA in the presence of sparse KGs. Our approach not only tackles the problem of graph incompleteness but also eliminates the common reliance on predefined neighborhoods for candidate answer selection, thereby overcoming a key limitation of previous methods. Extensive experiments conducted on multiple benchmark datasets demonstrate that our system substantially surpasses state-of-the-art approaches. These results underscore the effectiveness of our method in advancing multi-answer KGQA, particularly in scenarios where knowledge graphs are incomplete or sparse.