First person narrators can't know what's going on inside someone else.
But they can think they know.
And can narrate as if they know.
That's a key advantage of first person narrators. They can be unreliable.
I like your first example better... with some caveats.
The narrator presumes to know what is going on inside Tom.
That kind of "mind-reading" fantasy is common in us humans.
And it's a potentially interesting character flaw.
If you want us to think the character is flawed in that way,
this is fine way to illustrate the flaw.
The second narrator uses the word "seemed."
This makes me think the narrator is observing something.
But the narrator isn't telling us the observations, the sensory details that lead to "seemed."
Instead, the narrator summarizes the details, so that we can't share them.
This kind of summarizing is a potentially interesting character trait.
perhaps the narrator isn't aware of the sensory details,
but instead leaps to a guess about the feelings.
Being aware of guessing is a good thing.
But it will make it hard for the narrator to tell the story well.
If the narrator can't give us readers the sensory details,
it will be harder for us to share in the experience.
That's a problem for you.
If you don't mean to illustrate those character flaws
(mindreading, summarizing, or other form of unreliability),
it's probably better to have the narrator give us the sensory details.
What did the narrator see or hear
that led to those guesses about the other person's feelings?
Once the narrator tells us the sensory details,
the narrator might then conclude
something about the other person's feelings.
having seen both the details and the conclusion,
can then judge for ourselves
whether the narrator's conclusions seem reasonable.
Or the narrator might tell us the sensory details
and let us form conclusions
about the other person's feelings.