Confusing time stamps in TemporalFilter imshow function

Hello, maintainers, thank you for contributing this great work to BMTK.
I have some questions that I would like to ask for help. When I refer to the document (https://alleninstitute.github.io/bmtk/tutorial_filter_models.html) to learn how to use the filter model, I find that the timestamps displayed by the TemporalFilter imshow function are confusing. There are two graphs with similar response characteristics but different timestamps, and the one generated by the BMTK code is probably the wrong one because the response characteristics are even less similar to those in the paper.

paper

Durand, S., Iyer, R., Mizuseki, K., Vries, S. de, Mihalas, S., & Reid, R. C. (2016). A Comparison of Visual Response Properties in the Lateral Geniculate Nucleus and Primary Visual Cortex of Awake and Anesthetized Mice. Journal of Neuroscience, 36(48), 12144–12156. https://doi.org/10.1523/JNEUROSCI.1741-16.2016

Reproduce code

from bmtk.simulator.filternet.lgnmodel.temporalfilter import TemporalFilterCosineBump
tf = TemporalFilterCosineBump(weights=[30, -20], kpeaks=[50, 100],  delays=[0.0, 0.0])
tf.imshow()

Question

Is this a bug, should I use negative values ​​for timestamps?

Start in bmtk/simulator/filternet/lgnmodel/temporalfilter.py at TemporalFilterCosineBump.imshow(), and reproduce the supplied snippet. Compare how the displayed timestamps and response alignment relate to the linked paper; done means confirming whether negative timestamps are required and correcting the axis if the behavior is a bug.