Maybe the problem is adopt one clockwise and another counterclockwise?
This came to my mind when i attack the problem, but i went on just to see if i could try by this another way as well as adopt just clockwise [or counterclokwise]. But what i can't refut is why would it be wrong, that is:
## F*cos(wt) - [k(x1+x2)] - k(x1-x2) = m \frac{d^2 x1}{dt^2} ##
## [-k(x1+x2)] - k(x1-x2) = m \frac{d^2 x2}{dt^2} ##
the bracket being to the left spring:
If x2 = 0 and x1 > 0, will be a force on m1 in its negative direction, as to x2. Works as well to x1<0
Without bracket to the right spring:
If x2 = 0 and x1 > 0, will be a force on m1 in its negative direction, as to x2. Also works to x1<0
About the Latex, i will try ;)