You can solve both of your problems by setting the instrument to asynchonous mode. To do this, select the instrument, and click on the “Async” toggle button in the deck.
The reason why the AHDSR modulator’s attack period plays when the playback position loops is that repositioning the playback position on synchronous instrument retriggers that instrument. (This behavior is only rarely noticed, as it is only obvious in a rare few circumstances, such as yours.) Synchronous instruments are not retriggered when the playback position is repositioned unless the playback position leaves and re-enters the instrument’s trigger region.
The reason why a synchronous instrument immediately falls silent when the playback position is moved off its trigger region is that synchronous instruments always play the part of the waveform currently under the playback position. When the playback position moves to a position that’s not over any part of the waveform, there is no waveform to play, and so the instrument immediately stops producing output. Asynchronous instruments do not have this behavior, and so will play out to completion (or until their AHDSR modulator release tails have finished, if subject to AHDSR modulation).
Note that setting an instrument to asynchronous mode means you will no longer be able to seek within the audio files associated with that instrument simply by repositioning the playback position. It is still possible to seek within the assets by automating the instrument’s start offset property, but this can be a somewhat fiddly process.