Financial research reports require structured analysis, multimodal presentation, and timely incorporation of external information. Automating this workflow remains difficult because current systems struggle to coordinate long-form reasoning, gather reliable evidence at scale, and generate consistent visual outputs. To address these limitations, we propose ReportFlow, a multi-agent system designed to automate the end-to-end creation of macroeconomic, industry, and company research reports. ReportFlow focuses on improving workflow coordination, information retrieval, and multimodal generation, enabling the system to operate in a manner closer to real-world research teams. It consists of three key components. The Dual Graph Multi-Agent Architecture organizes the overall workflow: a Parent Graph performs global planning, while multiple Sub Graphs generate report sections in parallel, ensuring structured and coherent content. The Enhanced Hybrid Retrieval Augmented Generation module integrates sparse retrieval, dense retrieval, and real-time web search, allowing the system to gather comprehensive and current evidence. The HTML-to-Image generation module converts model-generated HTML into rendered figures, producing consistent and high-quality charts. We evaluate ReportFlow on a custom dataset of professional financial reports and observe substantial improvements over baseline systems in factual accuracy, analytical completeness, and multimodal quality.