For whatever reason Encylopedia Astronautica and wiki give the Ariane 1 a lower value for payload to LEO vs. GTO (1400 vs. 1850), which is highly odd. I can't think of any reason why that would be the case (GTO requires much more delta v than LEO). Going by the payload fraction to GTO, I get 1850/207200 = .0089, vs. 6300/705000 = .0089 for Proton-M, and 14220/738960 = .0192 for Delta IV Heavy. I think this is explainable by Ariane 1 being an older rocket with somewhat less efficient engines (the Viking-2s on Ariane 1 have 248 sec. sea level isp vs. 285 on the RD-253 and 365 on the RS-68).
EA gives the LEO performance for Ariane 5 as 16000 kg, but to a 51.5 degree inclination orbit, which is going to cut into the performance a lot. Arianespace's site gives a payload to LEO of 20 tons, which gives a payload/mass fraction of .0257, which is better but still not great. Payload to GTO is 10000/777000, which gives a mass fraction of.0129. Also, Ariane 5 does a bit of the same thing STS does and uses a hydrolox first stage with solid boosters. All the Ariane variants look like they are optimized for launches to GTO/GEO, so their payload fraction to LEO is going to suffer a bit. This explains why Ariane 4 and 5 have lower payload fractions to LEO (as far as I can tell your Ariane 4 numbers are correct).
To build on my previous calcs;
STS gets hurt a bit in the payload/mass calcs since it reuses so much of what it lifts to orbit. A better comparison would be Shuttle C, which likewises uses hydrolox/solids, and has a payload/mass fraction of 77000/1966675 = .0391, which is comparable to Delta IV Heavy.
Atlas V 552 (actually any Atlas variant without "0" in the middle digit) has solids in the first stage, which, if my thesis is correct, will lower the payload fraction. Theoretically the best Atlas V config for payload/mass fraction would be the Atlas V 402. Performance data for it can be found in the archived Atlas V Handbook.
Adding components from page 25 of that doc, we get a gross mass of 331702 kg for Atlas V 402. Wiki says that 402 has a payload to LEO of 12500 kg, which gives a payload/mass fraction of .0377, which is better, but not by much compared to 552. Interesting.