Hello Miguel,
I'm not one of the experts, but I'll give you my comments on what you're reporting and what I think I see in your Debug log.
I think the first thing to do is install the latest PHD2 development snapshot v2.6.11dev6), which you'll find here:
https://openphdguiding.org/development-snapshots/ The ASI174 guide camera you have should be running in 16bit mode (Camera Tools button in the Connect Equipment interface). In the Brain/Camera tab, set the Star Saturation Detection ADU to 65500 and not 255 as indicated in your log. Almost all the stars PHD2 had to work with were likely saturated at 2.5s exposure and ADU255 Max. The guide star flux density display should show a pointed profile and not the flat-topped ones you have been getting. Advice on the camera settings is given in the User Guide and Best Practices literature.
You comment that PHD2 always selects 'bad' stars when you can see 'good' stars on the display. This is a subjective response to the bad settings being used, PHD2 always selects the most appropriate stars from what is gets from the exposure (try 1-2sec), even though you would select something different. Let PHD2 make the decision to get the best outcome. If you can't see the selected stars in the camera display, move the gain slider to the left. Ensure it isn't selecting hot pixels by creating a Bad Pixel Map before starting the Calibration, etc.
So, run the latest PHD2 dev 6 version, do a Calibration using the new Calibration Wizard and then a 10-15min Guiding Assistant run, upload both Debug and Guide logs with your questions. Use the automatic log uploader provided in the PHD2 Help menu. The experts will then be able to tell you a lot more about your setup and how to get better results.
Cheers,
- Jack T