Feedback requested: improving inneffective nil coalescing diagnostic feedback

Development Compiler
10
With this test harness that mimics `??` and `flatMap` 
infix operator ??? : NilCoalescingPrecedence

// Edit: added this overload
@_transparent @_alwaysEmitIntoClient public func ??? <T: ~Copyable>(
    optional: consuming T?,
    defaultValue: @autoclosure () throws -> T // FIXME: typed throws
) rethrows -> T {
    switch consume optional {
        case .some(let value): value
        case .none: try defaultValue()
    }
}

@_transparent @_alwaysEmitIntoClient public func ??? <T: ~Copyable>(
    optional: consuming T?,
    defaultValue: @autoclosure () throws -> T?
) rethrows -> T? {
    switch consume optional {
        case .some(let value): value
        case .none: try defaultValue()
    }
}

extension Optional {
    @_alwaysEmitIntoClient public func myFlatMap<E: Error, U: ~Copyable>(
        _ transform: (Wrapped) throws(E) -> U?
    ) throws(E) -> U? {
        switch self {
            case .some(let y): try transform(y)
            case .none: .none
        }
    }
}

let optional: Int? = 1

let a = (optional ?? 10).flatMap { _ in 0 }    // 🔶
let b = (optional ?? 10).myFlatMap { _ in 0 }  // 🔶
let c = (optional ??? 10).flatMap { _ in 0 }   // ✅
let d = (optional ??? 10).myFlatMap { _ in 0 } // ✅

I am getting the warnings for those:

let optional: Int? = 1

let a = (optional ?? 10).flatMap { _ in 0 }    // 🔶
let b = (optional ?? 10).myFlatMap { _ in 0 }  // 🔶

but not for these:

let c = (optional ??? 10).flatMap { _ in 0 }   // ✅
let d = (optional ??? 10).myFlatMap { _ in 0 } // ✅

Which suggests there's some difference between my custom ??? operator and real ?? operator.. but what exactly is that difference and could ?? be modified to not have such behaviour?