yeah, I'm totally surprised, too.

(made an update to the previous zip) But I really need a break here.
Still some spoiled meshes, with
er045_0001_02_m01.mbg for example.
You may check out for yourself (using
hex2obj for example) which correction might be false;
or maybe vBlock and FIblock pairs don't fit together.
(abbreviations: vCnt= vertex count; FIcnt= face indices count)
# start of vertices
#0 59e
# UV01 0x7859c (2574) (name is uv01, but appears to be a vertex block for me)
#1 b6517
#2 c30de
#3 d202d
#4 e933a
#5 13cd61
# UV01 0x1db03f (2574)
#6 233baa
#7 247c2c
#8 270d6e
# FIs at:
# 0x70819
# 0xb36c9
# 0xc1d74
# 0xc3783
# 0xd12d9
# 0xe8058
# 0x10a80d
# 0x1acfdc
# 0x21616c
# 0x23f407
# 0x25dc57
# 0x292241
# SubMesh count and FIs block count are different!
Some SM may be missing. 11 != 12
#0 vCnt: 7461 (at 0x496), FIcnt: 21738
correction counts: 4887 / 16002
#1 vCnt: 502 (at 0xb6413), FIcnt: 2292
corr: FIcnt 5736
#2 vCnt: 613 (at 0xc2fd6), FIcnt: 1557
corr: FIcnt 2292
#3 vCnt: 959 (at 0xd1f29), FIcnt: 2226
corr: FIcnt 42
#4 vCnt: 1451 (at 0xe9236), FIcnt: 6603
corr: FIcnt 1515
#5 vCnt: 7461 (at 0x13cc59), FIcnt: 21738
corr: 4887 / 2226
#6 vCnt: 502 (at 0x233aa6), FIcnt: 2292
corr: FIcnt 6603
#7 vCnt: 959 (at 0x247b28), FIcnt: 2226
corr: FIcnt 16002
#8 vCnt: 1451 (at 0x270c6a), FIcnt: 6603
corr: FIcnt 5736