question about stages and prognosis

Survival rates are based on the original diagnosis.  If stage 1 had a 90% 5-year survival rate, that means that 90% of the people originally diagnosed as stage 1 are alive 5 years later.  The deaths include death from any cause, cancer, heart disease, auto accidents, etc.  Studies also often report "Cause-specific survival".  For that measure only deaths related to cancer (including death from complications of the cancer or the treatment) would be counted as a failure (or cause-specific death). 

When studies are analyzed for 5 year survival, some cases may not have follow-up data for 5 years.  Those cases are censored (dropped from the analysis) after the last time point where data is available for the case.

It depends on what you are looking for and it sounds like you should look for cause-specific survival.  There are other endpoints that are commonly reported, such as recurrence-free survival (death or any recurrence counts as a failure) or metastasis-free survival (death or distant metastasis counts as a failure but not local recurrence).

Some of the issues involved in reporting and choosing among different endpoints are:

1. It can be ambiguous whether or not to attribute a death to breast cancer.  If the person has breast cancer and heart disease and diabetes, for example, and dies of a heart attack or congestive heart failure, it seems like the death should be attributed to heart disease.  But the breast cancer and treatment may have weakened the patient and contributed to the death.
2. With early stage breast cancer most patients will live many years even if they eventually die of the disease.  In testing treatments researchers want to get answers about the success or failure of particular treatments and look for endpoints that can be analyzed sooner because waiting 15 years or more for an answer leaves them without an answer for many patients and with many more possible new treatments to consider and test scientifically.
3. For some treatments local recurrence is the best endpoint because the treatment is intended to reduce the likelihood of local recurrence (which ultimately should reduce the likelihood of distant recurrence and death).  Surgery and radiation therapy are treatments intended to reduce local recurrence, while chemotherapy and hormonal therapy are systemic and intended to reduce distant recurrence.