This bug may be fixed like bug #47858 (wrong line joins)

Yes, looks like it. That would be bug #32980. Ok to close as a duplicate unless there is some interest in keeping this open.

Example workaround, decimating the series by a factor of 10:

t = linspace (0, 1.2, 2^13);
y = min (1, 50 * exp (-10 * t));
noise = 0.001 * randn (size (t));
y = y + noise;
plot (t(1:10:end), y(1:10:end));
print ...

Dear developers

When the figure below is shown on screen it is correct, when printing to .pdf, .eps or *.ps the noise level is to high.

Problem only shows when using a large data set.

If there is no initial noise, there is no problem.

Problem does show with graphics_toolkit qt and fltk,
it does not show with gnuplot.

The problems shows on two different Linux hosts, with versions 4.0.0 and the developer version of Octave.

graphics_toolkit
% test PDF generation, qt and fltk produce wrong result, gnuplot is OK
t = linspace(0,1.2,100);  % no problem in PDF
t = linspace(0,1.2,2^13); % problem in PDF
y = min(1,50*exp(-10*t));
noise = 0.001*randn(size(t)); % no noise, no problem
y = y + noise;
MaximalAmplitude = max(y)

plot(t,y)
print -dpdfwrite TestPDF.pdf
print -depsc TestPDF.eps
print -dps TestPDF.ps