Aims. A spectroscopic analysis of stellar spectra can be carried out using multiple approaches such as different methods, line lists, atmospheric models, atomic parameters, and solar abundances. The resulting atmospheric parameters from these choices can vary beyond the quoted uncertainties in the literature. We characterize these differences by systematically comparing some of the commonly adopted ingredients such as line lists, equivalent width measurements, and atomic parameters. Methods. High-resolution and high signal-to-noise (S/N) spectroscopic data of one helium-core-burning red giant star in each of the three open clusters, NGC 6819, M67, and NGC 188 were obtained with the FIES spectrograph at the Nordic Optical Telescope. The M67 target was used to benchmark the analysis, as it is a well-studied cluster with asteroseismic data from the K2 mission. For the other two clusters we obtained higher quality data than had been analyzed before, which allows us to establish their chemical composition more securely. Using a line by line analysis, we tested several different combinations of line lists and programs to measure equivalent widths of stellar absorption lines to characterize systematic differences within the same spectroscopic method. Results. The obtained parameters for the benchmark star in M67 vary up to ~170 K in effective temperature, ~0.4 dex in log g and ~0.25 dex in [Fe/H] between the tested setups. Using the combination of an equivalent width measurement program and line list that best reproduce the inferred surface gravity from asteroseismology, we determined the atmospheric parameters for the three stars and securely established the chemical composition of NGC 6819 to be close to solar, [Fe/H] = −0.02 ± 0.01 dex. Conclusions. We highlighted the significantly different results obtained with different combinations of line lists, programs, and atomic parameters. The results emphasize the importance of benchmark stars studied with several methods to anchor spectroscopic analyses.