Large Language Models are legal but they are not: Making the case for a powerful LegalLLM