When analysing the influence of shade on photovoltaic (PV) systems, the intuitive assumption is that conversion architectures with more inverters, such as string-and module-inverters, will have lower energy losses associated with partial shading of the generators. However, other phenomena which can affect a system's response to shade, for example, the way in which the inverter performs the maximum power point tracking (MPPT), mean that modular configurations are not always the least susceptible to shade effects. We present a case in which the general tendency of many of the currently available commercial inverters to remain in local power maxima results in a string-inverter configuration behaving worse in response to shade than a central-inverter configuration. Experimental data were collected throughout 2005 and 2006 from PV plants in Arguedas and Sesma (Spain). We interpret our observations with theoretical analysis based on the results of simulation experiments.