You’ve made a slightly subtle, but common error. Observe that your “family of circles equation” omits $S_2$ itself: there’s no value of $\lambda$ for which $S_1+\lambda S_2$ is a multiple of $S_2$. That’s going to be a problem if $S_2$ passes through the point $(1,1)$, which it does when $p^2=6$.
You should instead start with the affine combination $(1-\lambda)S_1+\lambda S_2$, which doesn’t suffer from this flaw. It does omit the combination $S_1-S_2$, but that’s OK because that combination produces a line instead of a circle, anyway. If you work through the same calculations as before, you’ll end up with $\lambda(p+1)^2 = 2p+7$, which gives you a unique $\lambda$ unless $p=-1$, in agreement with the given solution.
That aside, I disagree with the given solution. The two circles only intersect when $p$ is in the (approximate) intervals $[-8.469,-1.744]$ and $[0.876,5.336]$†. Outside of these intervals there are no intersection points $P$ and $Q$ in the first place, so unless you’re allowing for imaginary intersection points, I would say that there are many more values of $p$ besides $-1$ for which there is no circle through $P$, $Q$ and $(1,1)$. Even taking this into consideration, though, $p=\pm\sqrt6$ do lie within the region in which the circles intersect, so your solution is still incorrect.
† More precisely, the endpoints of the intervals are the roots of $p^4+4p^3-44p^2-44p+69$, which are the values of $p$ for which the circles are tangent.