Question: Calculation of eigenvalues

Hello,

On a mechanical system that I could put in the classic form M*diff (x, t, 2) + K*x (t) = 0, I would like to calculate the angular frequencies.The methods that I know are:
- Calculate the solutions of the equation det (K-omega ^ 2 * M) = 0
- Or calculate the solutions of the equation det (M-1K-omega ^ 2 * I) = 0
- Or directly determine the eigenvalues ​​of M-1K using functions already created in Maple for example.

My problem is that with these three different methods, I get different results.The results of the first two methods seem to be close but still far from the third method (of course paying attention to the fact that the angular frequency is the square root of the eigenvalue).

Do you have an explanation? and what method allows me to have the good frequencies of my own system? I've heard about "stiff" problem that is to say problems with high values ​​of stiffness of the system which can generate numerical problems. This may be the case with the matrix K.

Here the calculation. 

[IMG]http://imageup.fr/uploads/1369151662.png[/IMG]

Thank you for your help.