Choosing significance level alpha and max_lag for Granger causality

I am really new to Granger Causality and am just rabidly experimenting with this MATLAB function:

granger_cause(x,y,alpha,max_lag)


http://www.mathworks.com/matlabcentral/fileexchange/25467-granger-causality-test

I have two series x and y which I need to check for causality. I need to check if y at a lag of 180 samples has a causal relationship with x or not.

Based on my experimentation, I noticed that if I vary the values of alpha and the lag, then I am able to find that almost anything seems to correlate with x (such as a random set of data). Can someone tell me just how should I should vary the alpha and lag values?

I found that if I put alpha as 0.8 then I get a good answer for x and y but not for the random data; however considering that other resources tell me to have alpha at 0.01 or so, I'm pretty suspicious about this. Can someone please help me out here?