Strange thing in Maxima: reduction of fraction

Hello everybody!

       I've run across such an odd thing:

(%i1) (p - q) / abs(p - q)^2;

                                         p - q
(%o1)                              --------
                                                  2
                                        (q - p)

(%i2) (q - p) / abs(q - p)^2;

                                            1
(%o2)                                -----
                                          q - p

Obviously, the output of (%o1) is correct, because I've not made any 
assumptions about p & q in advance. However, why did Maxima produce the 
output of (%o2) when I swapped the order of p & q as in the above?

Thanks a lot!