'Empty' ROC curve for decision tree

I created a spam/ham filters for emails with Matlab (data from UCI Machine Learning Repository: SMS Spam Collection Data Set). I used both Naive Bayes and Decision tree. I wanted to plot in a single graph the ROC curve. Following MathWorks I was able to build one ROC for Naive Bayes. However, with Decision Tree I have a few problems. This is the code I tried:

[respDT,scores] = resubPredict(Mdl);
diffscores= scores(:,2) - scores(:,1)
[X,Y,T,~,OPTROCPT,suby,subnames] = perfcurve(respDT,diffscores,1);
plot(X,Y)
hold on
plot(OPTROCPT(1),OPTROCPT(2),'ro')
xlabel('False positive rate')
ylabel('True positive rate')

And this is the image I get:

Now, as for the variables:

• Mdl = fitctree(X,Y) and its accuracy is more or less 94%
• Mdl.ClassNames is [0, 1]
• respDT is a vector with 0 (ham) and 1 (spam), 5417x1 double
• scores is 5417x2 double
• I created diffscores following https://uk.mathworks.com/help/stats/perfcurve.html

I am completely new to Matlab. Thank you for your help!