By prediction it is not necessarily meant that it must be a prediction of the future. In fact, this is very rarely the case. Scientific predictions are statements about what will be found if a certain observation is made. For example evolution predicts that if you examine the fossil record extensively enough, you will find creatures which share characteristics of mammals and reptiles. As far as predicting what man will look like one million years from now (I assume you meant homo sapiens, since homo erectus is extinct), I am pretty sure it can't be done with current scientific knowledge. We don't know what environmental changes will occur over the next million years. We don't know if there is sufficient variability currently in the human gene pool or if suffienct variability will arise to deal with any changes that might occur. It is possible that we will be much as we are today if there is little change in the environment (or if we deal with changes via technology rather than by natural selection.) It is possible that humans will be extinct in a million years. Nobody has ever said that the path taken by evolution is predictable in advance.
Not exactly a brilliant prediciton. Hell, I can predict I'll find the same living creatures with the same similarities and differences if I care to stroll about the planet a bit today. Haven't seen a dinosaur yet, but as far as I know mankind is spread fairly thin on the planet. There might be one out there. (I always check under my bed at night, just in case.)
It's the process at issue. Not the status quo.
So evolution boasts of consistent, natural process for millions of years, then pulls the rug out from under all history that may follow. Man, you can call that science if you want to, but I won't.