your data has zeros that make the mean closer to the small peak.
you should ignore the zeros, maybe you can throw them away OR you can try setting them equal to the non-zero data mean.
load('testdatarmoutlier.mat')
ind0 = y==0; % where are the zeros
meannz = mean(y(~ind0)); % mean of non-zero data
figure;
plot(x,y,'rx')
yline(mean(y),'k',Label='mean (close to bottom!)')
yline(meannz,'k',Label='mean ingnore zeros')
y(ind0)=meannz; % make zeros equal to the non-zero data mean
[y2,ind]=rmoutliers(y,'mean');
y(ind0)=0; % restore zeros
figure;
plot(x(~ind),y(~ind),'x')
