Despite the apparent simplicity of species distribution modelling approaches, the reliability of their predictions depends on the application of a number of good practices regarding the input data and the parametrisation of algorithms. In the context of invasion biology, inadequate modelling procedures may lead to erroneous conclusions regarding the potential spread of introduced species. However, clear guidelines for implementing these recommendations are often lacking, confusing or simply unknown by non-modeller endusers. Here, taking as an example the introduced land planarian Obama nungara, I fitted MaxEnt models applying six recommended processing steps with respect to sampling bias, predictor choice, training area, evaluation and hyperparameter tuning, separately or implemented together. I compared the resulting outputs to a model fitted with all default settings. All models differed from one another and from the default model, highlighting the importance of considering all these parameters when fitting species distribution models. However, the model that incorporated all fine-tuning methods was by far the most dissimilar, predicting much larger suitable areas globally, including in Africa where O. nungara has not been found so far. A closer examination suggested that it is likely a result of lower overfitting. This is a demonstration that modelling settings matter a lot, to the point that fined-tuned or default models may lead to considerably different conclusions when applied to invasive species.