It's not simply that drugs have failed time and again for Alzheimers. It's a very hard disease to come up with a treatment for. Here are some challenges:
1) It's a hard disease to detect in the earliest stages, where preventative/slowing drugs would have the most effect. If your drug slows decline by 30% but you start using the drug after 60% of the decline has already occurred, your effect size will be much smaller.
2) The blood-brain-barrier is a strong obstacle in our anatomy that prevents the passage of many potentially useful small or large molecules from general circulation into the brain.
3) Our early-stage disease models, things from petri dishes to mice, are less effective for choosing winners in this disease in particular. Maybe due to the complexity of the human neurology, though TBH I'm not sure why this is.
4) Even if you reduce beta-amyloid or other biomarkers, if you can't slow the disease progression/symptoms, then why is your drug useful? It's much harder to provide the slowing or reduction of cognitive decline than it is to simply clean up the beta-amyloid plaques. 4) is a specific case of the general use of "biomarkers" or "surrogate endpoints" - you need to measure SOME outcome, it's hard to do for cognitive diseases, so you find something measurable and look at that, though the linkage to symptoms is not totally clear in many cases.