To me, in your first image in this latest reply (Capture.PNG) the red lines on the DEM show a very good match compared to the first image you posted with the red lines on the ESRI topo.
There is one obvious mismatch where it looks like you drew a blue line. Possibly this is a very flat plain where the DEM has poor definition.
To can train the DEM to follow an actual creek or river, or to make it flow correctly "through" a road embankment at a road crossing, you need to "burn in the streams". The process to follow is to draw lines in a dataset called "agreestream" where ever you need to force a stream to be. This line should start and end on DEM cells that are correct to make sure there is good "continuity" between where the DEM gives you right answers and where you need to correct it. You don't have to put in lines for every stream in the DEM. You only need agreestream lines where the DEM is not "hydrologically correct". You can't always see this so it is a trial and error process.
Then you run the DEM Reconditioning with the agreestream dataset which lowers the DEM elevation where the streams need to be. Then build walls if needed. Building wall works the same as burning streams using a dataset named "innerwalls", but it results in a DEM where the cells are raised to stop flows from going where they shouldn't. Then fill sinks. Don't worry if the streams are burned really deep (-1000) or walls high (+500). The fill sinks will raise the DEM enough. The walls will still be high, but in the end you are not needing the right elevation, you are needing the right flow direction. You would always use the original DEM for elevations any way. The final agreeDEM will be "hydrologically correct".
After you fill the sinks the DEM should be "hydrologically correct" and the flow direction>flow accumulation and later steps should result in what you want. If not, add/adjust agreestream lines or innerwalls and repeat.
The highlighted steps in this screen capture of the ArcHydro menu shows the steps I use up to Stream Definition. I'm not sure if you are going beyond that.
Years ago (2012!) started a thread call "ArcHydro Problem Solvers" where I documented what I learned about ArcHydro "the hard way". Others added value to this and it is a good resource. The process I've used described there: https://community.esri.com/t5/water-resources-questions/archydro-problem-solvers/m-p/499209#M2440
Good luck!
Mark