Many data-intensive applications handle tabular data with more advanced structuring and processes than spreadsheets, enabling end-users to copy and adapt tabular data and processes to create new templates or datasets anytime. Recent research advances demonstrated that, in such clone-and-own scenarios, actions performed on the data structure, together with cloning and adaptation actions, can be captured within an operation-based model to prevent the drift of the internal tabular data model. However, this approach is limited by the assumption that each operation must maintain consistency regarding dependencies generated by the domain-specific languages that connect the observed and computed data. To address this challenge, this paper first introduces an evolved operation-based model that is designed to capture inconsistent tabular data while keeping a fine-grained trace of what part of the model is inconsistent. We then define specific trace operations to either fix a dependency in a model or remove one if its creating process is no longer relevant to the user. These operations support high-level editing scenarios on the tabular data, which enables easily fixing the equivalent of a spreadsheet formula or a process statement, or making the user aware that some part of the model is inconsistent while it is cloned. Additionally, we report on a positive scalability experiment on the tracing of large tabular data models with inconsistencies.