The XML-in-RDBMS approach has several limitations, including loss of type fidelity when converting data from complex structures to a string format like XML. This makes it difficult to accurately reproduce the original structure in a generic way without custom coding logic. Additionally, XML has poor handling for arrays and attributes can be on different axes than nested tags, leading to confusion and disambiguation issues. Furthermore, features and capabilities of XML in a column vary widely across RDBMS, making it challenging to ensure consistent behavior. The architecture also requires excessive developer-maintained utility code, impairing the smooth operation of SQL-based tools. In contrast, MongoDB's flexible document model and query language designed for rich shapes make it a better solution for handling complex data structures.