This is because of differences between which phonetic allophones BrE and AmE select for the perceived /ɑ/ phoneme, partly because AmE has no phonemic length but BrE does, partly because they have an [a] vowel that we do not and which we may hear as [æ].
The Brits do this even with words whose origins lies well beyond the old Roman Empire.
Last week during Japanese Prime Minister Shinzō Abe’s historic visit to the Pearl Harbor, I was stunned to hear the BBC pronounce that gentleman’s surname [ˈæbeɪ]. This is quite close to how one says the English word abbey [ˈæbiʲ].
This wasn’t just one newscaster alone, either, but several. I think I am not alone in finding this bizarre, as our asker’s question illustrates. And I think I now finally know the answer.
Abe is simply /ɑbe/ in Japanese. To an American, that becomes [ˈɑbeɪ], with the only modification an end-of-word off-glide /j/ for the /e/ vowel. To us, can and Khan are a minimal pair between /kæn/ and /kɑn/.
The common trait of many speakers of Southern British English to hear one thing but pronounce another even though it is different tells me that these must be allophones in the same phonemic bucked to them but not so to us. The most likely culprit here is the trap–bath split* where /æ/ becomes for them /ɑː/.
The linked Wikipedia article also mentions that the so-called “short A” of bath, which we use [æ] for, is often [a ~ a̠] in the North but [ɑː] in the South.
Notice how none of those is our own [æ]. Our own ears are like to throw all of [ɑ ~ a ~ ɐ] into /ɑ/ for we have no [a] at all.
So I think what’s happening is that the British are reproducing phonemic /ɑ/ as [a] or [æ] because it has no long marker for them as arbor or father would. We don’t do phonemic vowel length in North America, so we are not conditioned to make that sort of phonetic swap.
Because words like taco have no length marker as a hypothetical tarco would, they choose a short vowel there, and for them that’s if not exactly an [æ], at least something we perceive as such. I’m sure they don’t even realize they are doing this.
I think at all is a good illustration of this, because the unstressed /æ/ in at becomes [ɐ] or [ə]. So these Brits hear something they perceive as [ɐ] and therefore render as [æ] under stress, as in apple.
American speakers never do this, so it is bizarre for us to hear.