that is exactly what it is supposed to be, as your mediator ist negatively associated with your outcome. Thus, X has an indirect negative effect on Y mediated via its positive effect on M. So if X was manipulated experimentally you could maybe say, X reduced Y indirectly as it increased M. Just be careful with causal interpretation as this depends on your design. Also, you should look at the direct and total effects of X on Y as this will be more informative regarding the overall picture. It is possible to have a positive total effect of X on Y, and a negative indirect one via M. There could be other mediators at play (that you did not measure) that would then explain the positive indirect variance of X in Y. I would recommend reading Andrew Hayes: http://processmacro.org/index.html.
that is exactly what it is supposed to be, as your mediator ist negatively associated with your outcome. Thus, X has an indirect negative effect on Y mediated via its positive effect on M. So if X was manipulated experimentally you could maybe say, X reduced Y indirectly as it increased M. Just be careful with causal interpretation as this depends on your design. Also, you should look at the direct and total effects of X on Y as this will be more informative regarding the overall picture. It is possible to have a positive total effect of X on Y, and a negative indirect one via M. There could be other mediators at play (that you did not measure) that would then explain the positive indirect variance of X in Y. I would recommend reading Andrew Hayes: http://processmacro.org/index.html.