Strange behaviour on board editor/analysis board

If you set up the following position in the board editor:
twitter.com/ppamment/status/1043554686882713600

Note that this is setup as a standard rules game with black to move.

Then the analysis board offers kc8 as the best move, followed by rc1 for white, giving an evaluation of +0.9. Playing the fairly obvious rxd3 for white instead which is not offered as any of the top 5 suggestions, but when played the analysis shows +6.6.

Even stranger is that the third move suggestion for black is a kingside castle which is obviously not possible as the king is on d8 and the king rook is on e8.

This is reproducible at present at least for me.

You set the board up wrong, which lichess already told you on twitter
lichess.org/analysis/standard/r2kr3/pp6/2p4p/4B1p1/8/P2b4/1P2RPPP/3R2K1_b_-_-

Did you set this up with board editor or edit the FEN? They suggested I setup a 960 game, which I have tried again and verified it comes out like this even when the rules are set to standard so this does not appear to be the case.

Standard FEN import does not reject chess960 positions.

IMHO the question is why board setup didn't take you to lichess.org/analysis/chess960/r2kr3/pp6/2p4p/4B1p1/8/P2b4/1P2RPPP/3R2K1_b_-_- when you set up a chess960 position (from the Standard board editor, which maybe should have immediately set the mode to chess960 by giving Black castling rights).

#3 I used the board editor and cleared the castling rights checkboxes. Maybe you forgot to do that.

OK, yeah I assumed even if you left the castling checkboxes checked on the board editor it would automatically ignore castles that were no longer legal. I was wrong,
Perhaps it should though? I guess this is more of a feature request than necessarily a bug.