Consider a circle $C$, an ellipse $E$ strictly within $C$ except for a tangent point $p$, and a point $b$ strictly within $E$. Consider a point $q$ on $E$, and a point $q'$ given by the intersection of the extension of $qb$ with $E$, such that both $q$ and $q'$ are distinct from $p$. Define the points $r$ and $r'$ as the intersection of $C$ with the extensions of $pq$ and $pq'$ respectively.
We are interested in whether the point $j$, the intersection of the extension of $pb$ with $rr'$, lies inside the ellipse.
We can show (algebraically using Mathematica page 1,page 2) that the point $j$ is a property only of $C$, $E$, and $b$, independent of $q$, as illustrated in the figure:
So we would like some way to characterize this property (whether $j$ is interior to $E$) of $C$, $E$, and $b$, geometrically without introducing $q$.
$j$ on

If you are interested in the locus of point $b$ when $j$ is inside ellipse $E$, you could reverse your construction: take any point $j$ on $E$, trace from $j$ any line intersecting circle $C$ at $r$, $r'$ (that line is taken parallel to $x$ axis in figure below), join $pr$, $pr'$ to find their intersections $q$, $q'$ with $E$. Point $b$ lies on segment $qq'$. Repeating the construction for another line $tt'$ through $j$ you can find a second segment, whose intersection with $qq'$ is then point $b$ (otherwise, you could also find $b$ as the intersection of $pj$ and $qq'$).
I didn't embark into a computation, but with GeoGebra I could find the locus of $b$ as $j$ varies on the ellipse: it should be another ellipse (pink in figure below), tangent at $P$ with $C$ and having its axes parallel to those of $E$.