Densely deploying Massive MIMO access points results eventually in cell-free systems. To achieve scalability, users should be allocated only to neighbouring access points as it is not realistic to assume that all access points serve all users. At the same time interference from other nearby users should be suppressed, as it is not practical to suppose that channel state information from all users is available for full MMSE-based interference suppression. Motivating the use of partial MMSE combining vectors. This work shows the taxonomy of proposed MMSE combining vectors for different clustering techniques, and three approaches to partially suppress interference based on 1) interferers with common access points, 2) strongest channel gain and 3) strongest eigendirections. In order to analyse the pros and cons of the three interference suppression methods, a general parameterised MMSE combining vector is proposed, that is generic enough to represent any clustering methodology and any partial interference suppression technique. The results are presented in terms of spectral efficiency and computational cost, with the aid of a dense indoor massive MIMO experiment and contrasted with an indoor simulation using the WINNER II channel model. Experimental and simulation results show that partial suppression based on shared access points is not suitable for systems with a small probability of users sharing the serving access points, as dominant interference is likely to come also from users connected to access points not serving the harmed user. Channel gain based interference suppression provides the best controllable approach in terms of spectral efficiency and computational cost. However, eigendirections-based interference suppression achieves the highest spectral efficiency, when the number of antennas per access point is high enough.