On simultaneous best L1 approximations in C[−1, 1]