This may be the first proposal I've seen to move from control-rail activation to infrared or contactor! Usually operators are going the other way, to get away from adjustment and reliability problems.
Yes, Ken, everything is working as expected, including the sparking. That happens whenever the connection through the axles, from wheels on one side of the train to the other, is broken. Ideally, this should happen only once, as the last wheel set leaves the control rail; so the dirty-rail or dirty-wheel diagnosis is correct; and cleaning wheels and rails (both the control rail and the other outside rail) should make a difference, although there will always be some sparks.
Another thing that might help is to see whether the sparking is greater on one side of the track or the other, and to swap the control rail to the other side if that might make the sparking less obvious on your particular layout.
A relay circuit would just substitute the inductive relay winding for the inductive semaphore winding, but probably wouldn't solve the underlying problem.
(There's no "positive" or "negative" in a circuit like this, nor in most toy-train circuits. One can call a toy-train frame a "ground," since the chassis of any vehicle is an honorary ground, even though it is not really grounded. But otherwise, as in the semaphore or the U terminal of a multi-control transformer, the proper term is "common.")