Unless you have a typo or have left out some other information it appears that your
models = \[Phi]s Exp[-Exp[0.92` (n - M)] + 0.92` (n - M) \[Alpha]]
n and M values are not independent. If you choose any value for n then M just needs to be adjusted to match.
This seems in agreement with your observation that the Fit works just fine if you fix the value of n.
This is like I tell you the equation for a line is
y = 3(n-M)+2, tell you y=7 and ask you what the value of n and M are.