Annotation times cannot be used to get corresponding data

Description of the problem

I have some (manually-created) annotations for a raw file. I want to pull out the data from a particular channel, from the start to the end of that annotation. Either I'm making a really dumb mistake, or this is surprisingly difficult to do. It's clear that the problem has to do with first_samp (and I could subtract raw.first_samp somewhere in my code to get the right result) but it feels like this should be easier / "just work".

Steps to reproduce

import matplotlib.pyplot as plt
import mne
import numpy as np

fname = mne.datasets.sample.data_path() / "MEG" / "sample" / "sample_audvis_raw.fif"
raw = mne.io.read_raw_fif(fname)
annot = mne.Annotations(onset=3, duration=1, description="test")
raw.set_annotations(annot)

ann = raw.annotations[0]
data = raw.get_data(picks=[0], tmin=ann["onset"], tmax=ann["onset"] + ann["duration"])[0]
plt.ion()
fig, ax = plt.subplots(layout="constrained")
ax.plot(np.arange(len(data)), data)

raw.plot(picks=[0], start=ann["onset"], duration=ann["duration"])

Link to data

No response

Expected results

1. when I call raw.plot(start=raw.annotations[0]["onset"], duration=raw.annotations[0]["duration"]) I would really like to see the data corresponding to that annotation span
2. when I call raw.get_data(tmin=raw.annotations[0]["onset"], tmax=raw.annotations[0]["onset"] + raw.annotations[0]["duration"]) I would really like to obtain the data corresponding to the annotation span

Actual results

NB: although sys_info() below says I'm on 1.11.0, I get the same results on main --- just happened to be in a project venv when I stumbled on this.

Additional information

Platform Linux-6.8.0-45-generic-x86_64-with-glibc2.39
Python 3.13.12 | packaged by conda-forge | (main, Feb 5 2026, 05:53:46) [GCC 14.3.0]
Executable /home/drmccloy/Documents/academics/research/ilabs/prism/.pixi/envs/analysis/bin/python3.13
CPU Intel(R) Core(TM) i9-9900K CPU @ 3.60GHz (16 cores)
Memory 125.7 GiB

Core
├☒ mne 1.11.0 (outdated, release 1.12.1 is available!)
├☑ numpy 2.4.2 (OpenBLAS 0.3.30 with 8 threads)
├☑ scipy 1.17.1
└☑ matplotlib 3.10.8 (backend=qtagg)

Numerical (optional)
├☑ sklearn 1.8.0
├☑ numba 0.64.0
├☑ nibabel 5.4.0
├☑ nilearn 0.13.1
├☑ dipy 1.11.0
├☑ openmeeg 2.5.16
├☑ pandas 3.0.1
├☑ h5io 0.2.5
├☑ h5py 3.15.1
└☐ unavailable cupy

Visualization (optional)
├☑ pyvista 0.47.1 (OpenGL 4.5.0 NVIDIA 595.58.03 via NVIDIA GeForce RTX 2060/PCIe/SSE2)
├☑ pyvistaqt 0.11.3
├☑ vtk 9.5.0
├☑ qtpy 2.4.3 (PySide6=6.9.2)
├☑ pyqtgraph 0.14.0
├☑ mne-qt-browser 0.7.4
├☑ ipywidgets 8.1.8
├☑ trame_client 3.11.3
├☑ trame_server 3.10.0
├☑ trame_vtk 2.11.1
├☑ trame_vuetify 3.2.1
└☐ unavailable ipympl

Ecosystem (optional)
├☑ mne-bids 0.18.0
├☑ mne-icalabel 0.8.1
├☑ mne-bids-pipeline 1.10.0.dev167+g18af7ca2d
├☑ eeglabio 0.1.3
├☑ edfio 0.4.13
├☑ curryreader 0.1.2
├☑ mffpy 0.10.0
├☑ pybv 0.7.6
├☑ antio 0.6.1
├☑ defusedxml 0.7.1
└☐ unavailable mne-nirs, mne-features, mne-connectivity, neo

Start by reproducing the provided script, then inspect the Raw.get_data and Raw.plot entry points for how they interpret annotation onsets alongside first_samp. Trace the resulting sample ranges and verify that both APIs select and display the data covered by the annotation without requiring callers to adjust the onset.