Mixed Model: Between Subjects (Continuous) Within Subjects (4 Levels)

[gheffo]

Hi everyone,
I'm trying to test a model with a between subjects variable (continuous) and a within subject variable (with 4 levels) in predicting a continuous dependent variable.
I used the GLM --> Repeated Measures (SPSS) technique posted by Karen (http://talkstats.com/showthread.php?t=5501) where I median split on the continuous between subjects variable and that gave me the results and graph I think I should be looking for. Is this a possible way to test the model?

Would it be more statistically sound to use the continuous between subjects variable as a covariate in the model (thereby avoiding the median split)? I tried this, but was unable to get the graph. Please let me know what you all think and if there is a better way to test this (e.g., regression)!?

Thank you,
Geoff