It's not like the default Android SMS app indicates that messages have been deleted. And there was no provenance information provided as to where or when they were collected.
I hate this, because screenshots come out looking like trash and it's very difficult with most messaging apps to show the timestamps for all messages. Eventually, this pendulum is going to swing (when cases start getting thrown out for this lack of timestamps/evidence of deleted messages, etc) and law enforcement (at least in the military environment) will have a bit more support in pulling relevant (and only relevant) data from victim devices for the purpose of evidence collection.
One additional thing about screenshots: They can be totally faked, and the "contact" can't be validated from conversation screenshots. For example, if you buy a burner phone, you can create a whole conversation as though the burner phone is the assailant, then change the contact's number to the real assailant. Some chat apps keep the whole conversation, despite the number change, in the same chat and make it impossible to tell which number sent the messages. Cellebrite indicates the number (assuming we're talking about SMS here) where the message came from, even if the contact changes.
The 3 points to Miranda rights are cops, custody, and questioning(provided the questions would reasonably illicit a criminal response).
For example, if I am a cop and I walk up to you on the street and ask you questions, your 5th amendment only applies if I'm not letting you go (custody). There's some ambiguity on the custody part, but it's usually revolving around the idea that you're under arrest. For the military, since you're always "Subject to the code(UCMJ)", the custody part of Miranda means that the only factors that matter for cops (and this is more broadly inclusive of authority figures or anyone who is a "mandatory reporter", meaning someone who must report criminal activity/statements to military law enforcement) and questioning (so, is the subject being asked questions that would elicit a criminal response). If I (a cop, commander, supervisor, whatever) I a subordinate a question that I know is likely to elicit a criminal response, that response (and the fruits of it) are inadmissible in court and cannot be used to further a criminal investigation. So, I can ask Private Dump "Hey Dump, what did you do this weekend?", and be fine, but if I ask "Hey Dump, how much Cocaine did you snort off that stripper's chest Saturday?", I would be eliciting a criminal response (assuming that I have reason to believe he was engaged in criminal drug activity on Saturday).
This tends to work in favor of the subject of an investigation, though, because they a) can't incriminate themselves without a rights advisement unless they willingly volunteer incriminating information in response to a question that wasn't likely/designed to elicit it and b) they have to waive their rights before asking these kinds of questions, so they know they are being investigated (and what for, since Article 32 advisements are WAY more thorough than Miranda).
As for your point about this and other amendments being mute for military, there are a couple ways to think about it. If a military superior or law enforcement conducts a search without search authorization or consent, the results cannot be used against you criminally, on administratively. Effectively, they could kick you out (but not with any of the negative flags, like a dishonorable), but they can't prosecute you. This applies to searches, asking questions without Art 32 advisements, etc. Effectively, your rights are still being respected and they're interacting with you like an employer, just with a bit more authority.
My preference would be a process whereby cops run the Cellebrite extraction, then produce a report that's limited to the content the victim and her counsel agree to, like conversations between her and Bob, and conversations between her and Sally, but nothing more. This way, at least we have some data that is verifiable and detailed (would show the to/from numbers, contact info, times, etc).
These are consent-based searches, not warrantless.