How to estimate transition probabilities in the absence of individual patient data?

What I tried so far is to read of the survival curves given in clinical trials with Plot Digitizer, with months ((t)months) as the X axis and probability (S(t)) on Y axis (0-1).

I then imported the digitized points to STATA and ran

stset tmonths, fail( St)

streg, d(weibull)

From the results I used:
Haz. Ratio of _cons as lambda and haz.ratio of /ln_p as gamma and got the transition probability for each respective cycle in Excel with the formula =EXP(lambda*timeinmonths^gamma).

I just wonder if this is the right approach as I feel I'm messing something up.