No.
You have to use it with some common sense.
Not just inputting PBs but what is a more realistic estimate of an anthlete's ability at the time.
From above.
45.5PB is not a realistic estimate for his London shape. That was an early season run & he's quoted that he doesn't do much speedwork for his February/March races meaning he ran that 45.5 off his endurance work.
It was probably about 45-flat in London ( maybe even into high-44s )
45.0 / 1'40.9 ->3'27.2
As a theoretical 1500 if his endurance lasted "perfectly" upto 1500 of just "pure" 800 training. It woudn't & anyone with a clue about him woud know he'd be doing damn well to even break 3'35, so a 3'27 potential is of little relevance ( It also predicts a 13'30 for a 5k for him. Is that of any relevance to him ?! )
Read above.
You put in an almost certainly wrong 400 for his London shape & he doesn't train for the 1500.
This is where you have to use some common sense.
I know nothing about you, but your 5 & 10k aren't of similar "quality".
Rule of thumb for an equally strong 5k/10k guy is
10k = 2*5k + 1'00
Your 5k & 10k times are mismatched & you will get nonsense 400/800 times.
I see you have given us a crucial extra bit of info & say you had a 56s for 400 ( assuming in same shape as the 5k/10k ).
Using 56 & the 17'30 & 36'00+ & guidance, a very rough prelim estimate for your then metric potential is
56.0 / 2'05.1 ->
4'16.9
5'56.1
9'22.8
16'38.4 !!
36'00.5
Are any of the 800, 1500, 2k or 3k close to what you have run in your ideal race ?
Looking at the 5k/10k there, it actually indicates your 5k of 17'30 is very poor & if you had 56 speed & 36'00 endurance at the time you shouda been running nearer 16'38 in your perfect 5k not a 17'30
A calculator doesn't know what is a sprint or endurance. You have to input realistic times & use common sense about an athlete's ability
Trying it out, the "NOT shorter than 400 looks accurate"
See line above with 56.0 & ~ 2'05.1
Without knowing anything about you apart from what you've said, I'd say you 800 was ~ 2'05.
Anywhere close ?
I've tried it a lot & if you use good inputs ( remember GIGO ) it's accurate from 400 - 10k ( & even M for Paula !!! )