This is what's called coarticulation: sounds are influenced by the surrounding sounds. When you say nicht Gräfinger, the /t/ is unvoiced, but the following /g/ is voiced. Your vocal tract cannot switch off the vibrating glottis that fast, so there this feature will bleed into the following sound: either the /g/ becomes unvoiced and changes to /k/, or the /t/ becomes voiced and turns into a /d/. What actually happens depends on the individual speaker I'd guess, though it seems more likely that the onset of the voiced consonant shifts forward, ie the /t/ would change to /d/.
Just bear in mind that your vocal tract is a physical system, which means there is a latency when its moving parts change position.
In Ahornstrasse the speaker indeed produces a sound at the morpheme boundary. As the tongue moves from the alveolar position of /n/ to the palatal /S/, it is just in the same position as for the /i/ sound, which then 'accidentally' gets produced.
In example 2 I can actually hear a glottal stop before the /g/, though the /t/ still sounds like a /d/ just before then.
In the Wie bitte example I would assume that as it is the initial sound, the glottis is not instantaneously producing the voiced carrier, and thus the /v/ turns into the unvoiced /f/ initially. If you were humming before saying it, the glottis would be vibrating, and then it is more clearly a voiced /v/ sound.