Listening to something and understanding something are two completely different things. I don't believe you processed them fully enough to train your brain. If you did, your music would be more refined. Just calling it like it is.
You say it's "easier said than done", which is true, but it's not an excuse to spend less time on it and assume you'll never get it. Assuming you'll never get it is a great way to not get it. So you have to put more of your time into learning how to write for orchestra if you're going to want to write something more realistic.
I do hear the similar styles, but you're still missing the expression present in the SMG example, real or not.
Your orchestra is a little distant, which means the close / room / hall mic mix is skewed towards the hall mic (or away from the close mic) or you don't have the flexibility to mix all three together (you wouldn't beneath Platinum version of EWQLSO I believe). Or your samples are pre-baked in reverb.
Your brass in particular is noticeably lacking dynamic crossfade automation, so the high-dynamic "blatty" tone to the brass is constantly there, without an emulated decrease in incoming breath via a lower dynamic (often done with CC1 or CC11). Even if you don't have dynamic crossfade, you could at least record volume event edits. It's not the same thing, but it approximates it.
You haven't fully accounted for the slow attacks of certain articulations (particularly in the non-staccato brass and non-staccato strings), so the slow articulations are late. Therefore, it would help to shift those slow articulations back in time a little to make sure they are more on-rhythm.
As Slimy mentioned, a lot of your notes sound quantized. One way to help that is to write bigger chords, and offset the notes in your chords. This at once gives you a bigger sound and a bit more flexibility when it comes to aligning the note transients.