Out of curiosity I took some guesses:
Normally you would model friction as ##f=\mu mg## for the cart ... is there friction in the pulley too?
m=mass of trolley, M=total mass, then hanging mass is M-m > m (##\mu## needs to be measured.)
Free body diagrams for trolley accelerated by hanging mass.
(1) ##T-\mu mg = ma## for the cart
(2) ##(M-m)g - T = (M-m)a## for the hanging mass
Solve for ##a##:
(3) ##a = g - (1+\mu) mg/M##
experiment:
- change mass by taking weights off the trolley and adding them to the hanging mass - so M is a constant.
- measure the acceleration (time T over a fixed distance d ##a=2d/T^2##)
expect the plot of ##a## vs ##m## will be a straight line with slope ##s=-g(1+\mu)/M## and intercept ##g##
(does this sound like what you did?)
This allows us to calculate the friction coefficient from the measured slope.
If ##f## decreases with ##m##, then ##\mu## decreases with ##m## a lot.
This will change the shape of the graph - so: plot your data - is the graph a reasonable straight line?
If not, then what sort of curve is it?
Are there systematic errors in your experiment?
ie. the above calculation neglects friction and mass in the pulley... was this a good idea?