looking at it strictly from geometry standpoint, canards might lead to a few more hotspots of rcs return compared to conventional layout, when looked at from the front. naturally, when in a chase, canard layout might offer less returns than conventional one, for the exact same reasons.
that's if canards/horizontal tail surfaces would be at the exactly the same level.
of course, having canards at the exactly same level as wings lowers their aerodynamic effectiveness. which is why j20 has them canted so their vortexes go over the wing.
so yes, j20 layout should lead to a few more rcs hotspots than j20 overall, both when looked from the front and the back. of course.
situation A) enemy radar is mostly in front of the j20/f22, both azimuth wise and elevation wise. leading surfaces of canards deflect to the side, no problem. leading surfaces of wings do the same. edges of canards, canard roots, slats on the wings and wing edges produce returns in all directions.
here j20 is potentially at a disadvantage because it has two canard edges, canard root edge, two edges from one of the slat/wing joints (other one is possibly covered by the canard) and the wing's edge. that's 6 sharp edges that might radiate in all directions. f22 in same situation would have two pairs of slat/wing edges plus a wing edge. its wings are obstructing line of sight to the tails so no edges are visible. that's 5 sharp edges per wing. (10 total for aicraft versus 12 for j20). BUT, since j20 canards are canted, it's possible the inner slat/wing joint isnt actually covered/obstructed by them. if so, that'd add 2 more edges per wing for a total of 10 versus 16.
situation B) enemy radar is front but is little below the level of the j20/f22. again, leading surfaces are not an issue. this time raptors tails' edges are exposed. so we have the same as above for raptor plus two horizontal tail edges plus tail root edge. thats 8 per side or 16 total. j20 this time has no chance of its canards obstructing any part of the wing so we're automatically talking about 8 edges per side or 16 total.
situation C) enemy radar is in front but is little above. this time j20 has theoretically a chance that its canted canards will obstruct one of the slat/wing joints so it may actually have 12 edges visible versus 16 on raptor.
but - one really designs all those shapes around the expected percentage of situations planes will find themselves in.
when one looks at a plane from the ground at 400 km away, one will have radar look at the plane at similar angle as situation A. as the plane moves closer, situation will change slowly to situation B. at which point does B become applicable? i don't know precisely. At 100 km away, a plane flying 15 km higher than the radar will be seen at an offset of some 6-7 degrees. that should be more than enough for a clear view of the horizontal tails. perhaps even half as large offset might be enough.
on the other hand, having platforms in the air that will look at j20 from above is hard. unless we're talking about awacs and a ground hugging j20. those situations will of course be much less common than a variety of ground radars searching for a high flying j20.
situation D - radar is at just the angle offset from the longitudinal axis of the plane so it can catch the big return of wings' leading surfaces. j20's surfaces are caught the same way as in situation A, while raptor still obstructs view to tails. but all that is less important since big return of whole length of leading surfaces dwarfs the combined returns of those tiny edges in situations A/B/C.
all this is very simplified, and there's a bunch more potential sources of return on the planes which i havent mentioned.
and none of this mentions materials, which may have progressed to be point where its practical to rely on them negating those edge returns for select radar wavelengths. so, shape wise, canards may be slightly worse solution, but in the end, when materials and angles at which planes get looked at and number of rcs spots are calculated - overall increase in rcs may not be even close to order of magnitude higher, and for certain wavelengths it may be basically the same.
.
